ORIGINAL RESEARCH published: 17 April 2019 doi: 10.3389/fnut.2019.00045

What’s Normal? Microbiomes in Human Milk and Infant Feces Are Related to Each Other but Vary Geographically: The INSPIRE Study

Kimberly A. Lackey 1, Janet E. Williams 2, Courtney L. Meehan 3, Jessica A. Zachek 2, Elizabeth D. Benda 2, William J. Price 4, James A. Foster 5, Daniel W. Sellen 6, Elizabeth W. Kamau-Mbuthia 7, Egidioh W. Kamundia 7, Samwel Mbugua 7, Sophie E. Moore 8,9, Andrew M. Prentice 10, Debela Gindola K. 11, Linda J. Kvist 12, Gloria E. Otoo 13, Cristina García-Carral 14, Esther Jiménez 14, Lorena Ruiz 15, Edited by: Juan M. Rodríguez 16, Rossina G. Pareja 17, Lars Bode 18,19, Mark A. McGuire 2 and Aldo Corsetti, Michelle K. McGuire 1* University of Teramo, Italy 1 Margaret Ritchie School of Family and Consumer Sciences, University of Idaho, Moscow, ID, United States, 2 Department of Reviewed by: Animal and Veterinary Science, University of Idaho, Moscow, ID, United States, 3 Department of Anthropology, Washington Arianna Aceti, State University, Pullman, WA, United States, 4 Statistical Programs, College of Agricultural and Life Sciences, University of University of Bologna, Italy Idaho, Moscow, ID, United States, 5 Department of Biological Sciences, University of Idaho, Moscow, ID, United States, Veronique Demers-Mathieu, 6 Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada, 7 Department of Human Nutrition, Egerton Oregon State University, United States University, Nakuru, Kenya, 8 Department of Women and Children’s Health, King’s College London, London, United Kingdom, *Correspondence: 9 MRC Unit The Gambia at the London School of Hygiene and Tropical Medicine, Fajara, Gambia, 10 MRC International Michelle K. McGuire Nutrition Group, London School of Hygiene and Tropical Medicine, London, United Kingdom, 11 Department of Anthropology, [email protected] Hawassa University, Hawassa, Ethiopia, 12 Faculty of Medicine, Lund University, Lund, Sweden, 13 Department of Nutrition and Food Science, University of Ghana, Accra, Ghana, 14 Probisearch, Tres Cantos, Spain, 15 Department of Microbiology Specialty section: and Biochemistry of Dairy Products, Instituto de Productos Lácteos de Asturias (IPLA-CSIC), Villaviciosa, Spain, This article was submitted to 16 Department of Nutrition, Food Science, and Food Technology, Complutense University of Madrid, Madrid, Spain, Food Microbiology, 17 Nutrition Research Institute, Lima, Peru, 18 Larsson-Rosenquist Foundation Mother-Milk-Infant Center of Research a section of the journal Excellence, University of California, San Diego, La Jolla, CA, United States, 19 Department of Pediatrics, University of Frontiers in Nutrition California, San Diego, La Jolla, CA, United States

Received: 20 December 2018 Accepted: 27 March 2019 Background: Microbial communities in human milk and those in feces from breastfed Published: 17 April 2019 infants vary within and across populations. However, few researchers have conducted Citation: Lackey KA, Williams JE, Meehan CL, cross-cultural comparisons between populations, and little is known about whether Zachek JA, Benda ED, Price WJ, certain “core” taxa occur normally within or between populations and whether variation Foster JA, Sellen DW, in milk microbiome is related to variation in infant fecal microbiome. The purpose of this Kamau-Mbuthia EW, Kamundia EW, Mbugua S, Moore SE, Prentice AM, study was to describe microbiomes of milk produced by relatively healthy women living at K. DG, Kvist LJ, Otoo GE, diverse international sites and compare these to the fecal microbiomes of their relatively García-Carral C, Jiménez E, Ruiz L, Rodríguez JM, Pareja RG, Bode L, healthy infants. McGuire MA and McGuire MK (2019) Methods: We analyzed milk (n = 394) and infant feces (n = 377) collected from What’s Normal? Microbiomes in Human Milk and Infant Feces Are mother/infant dyads living in 11 international sites (2 each in Ethiopia, The Gambia, Related to Each Other but Vary and the US; 1 each in Ghana, Kenya, Peru, Spain, and Sweden). The V1-V3 region Geographically: The INSPIRE Study. Front. Nutr. 6:45. of the bacterial 16S rRNA gene was sequenced to characterize and compare microbial doi: 10.3389/fnut.2019.00045 communities within and among cohorts.

Frontiers in Nutrition | www.frontiersin.org 1 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

Results: Core genera in feces were Streptococcus, Escherichia/Shigella, and Veillonella, and in milk were Streptococcus and Staphylococcus, although substantial variability existed within and across cohorts. For instance, relative abundance of Lactobacillus was highest in feces from rural Ethiopia and The Gambia, and lowest in feces from Peru, Spain, Sweden, and the US; Rhizobium was relatively more abundant in milk produced by women in rural Ethiopia than all other cohorts. Bacterial diversity also varied among cohorts. For example, Shannon diversity was higher in feces from Kenya than Ghana and US-California, and higher in rural Ethiopian than Ghana, Peru, Spain, Sweden, and US-California. There were limited associations between individual genera in milk and feces, but community-level analyses suggest strong, positive associations between the complex communities in these sample types. Conclusions: Our data provide additional evidence of within- and among-population differences in milk and infant fecal bacterial community membership and diversity and support for a relationship between the bacterial communities in milk and those of the recipient infant’s feces. Additional research is needed to understand environmental, behavioral, and genetic factors driving this variation and association, as well as its significance for acute and chronic maternal and infant health.

Keywords: human milk, breastmilk, feces, microbiome, international, infant, breastfeeding, maternal

INTRODUCTION by women living in Finland (n = 18), while Davé et al. (9) reported Streptococcus, Staphylococcus, Xanthomonadaceae, Although long thought to be sterile, human milk is now known and Sediminibacterium were the most abundant taxa in milk to contain myriad bacteria, and growing evidence suggests produced by Mexican-American mothers (n = 10). In Chinese that the composition and profiles of these microbiomes differ and Taiwanese women (n = 133), family-level analysis revealed among geographically distinct populations of women. The Streptococcaceae, Pseudomonadaceae, Staphylococcaceae, microbiome of milk produced by healthy women is of scientific Lactobacillaceae, and Oxalobacteraceae as the most abundant and public health interest because these microbes may, at least taxa (10). Other reports suggest additional differences among in part, determine which microbial communities are in the populations (11–16), although some similarities are notable, gastrointestinal (GI) tracts of their infants (1–5). The infant GI such as the dominance of members of the Streptococcaceae microbiome (often assessed through the analysis of feces) is of and Staphylococcaceae families. It is unknown, however, if this substantial interest because its variation has been associated with variation is due to genuine differences among populations, a variety of human diseases, both in early and later life [reviewed differences in location of milk collection (e.g., hospital vs. home), in (6)]. or differences in sample collection methods, storage, processing, In the first report of a complex bacterial community in human and analyses. This is a persistent problem in microbiome research milk using high-throughput methodology, the milk microbiome and can only be solved with rigorously controlled studies of of healthy women (n = 16) in the Moscow, ID/Pullman, WA representative cohorts of women from diverse populations, region of the United States was found to be dominated by including standardized milk collection protocols. Streptococcus, Staphylococcus, Serratia, and Corynebacterium To this end and to help address other potential confounders, (7). Bacterial communities appeared to be somewhat unique Kumar et al. (17) investigated the influence of geographic for each woman, although 9 “core” genera were common all location on the milk microbiome by collecting and analyzing samples. Since the publication of this paper, additional studies milk produced by 80 healthy women (20 each from Spain, have suggested that the primary bacterial taxa in milk vary Finland, South Africa, and China) at 1 mo postpartum. across populations (Supplementary Table 1). For example, Substantial differences were found among cohorts, and variation Cabrera-Rubio et al. (8) found that Leuconostoc, Weisella, and was related to a variety of factors, such as delivery mode. For Lactococcus were the most predominant genera in milk produced example, milk produced by Chinese women contained relatively more Streptococcus than milk produced by women in all other Abbreviations: ANOSIM, analysis of similarity; ANOVA, analysis of variance; GI, cohorts, while milk produced by Spanish women had relatively gastrointestinal; GLIMMIX, generalized linear mixed model; HMO, human milk more Propionibacterium and Pseudomonas than that produced in oligosaccharide; ASV, amplicon sequence variants; ETR, Ethiopia, rural site; ETU, other locations. This study also demonstrated that milk produced Ethiopia, urban site; GBR, The Gambia, rural site; GBU, The Gambia, urban site; GN, Ghana; KE, Kenya; PE, Peru; SP, Spain; SW, Sweden; USC, United States, by Spanish and South African women was characterized by California site; USW, United States, Washington site. relatively higher proportion of bacterial genes involved in lipid,

Frontiers in Nutrition | www.frontiersin.org 2 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study amino acid, and carbohydrate metabolism than that of Finnish each study location. Sample collection took place between women. As such, not only do there appear to be genuine May 2014 and April 2016 and was carried out as a cross- compositional differences in the milk microbiome around the sectional, epidemiological, multi-cohort study. Informed written world, but there may also be differences in microbial function. or verbal consent was obtained in the local language from each Similarly, a substantial and growing literature exists regarding participating woman in her primary language. Informed verbal the human fecal microbiome during infancy [for example, (2, consent was obtained when a subject’s literacy level prevented 18–22)]. These studies also report differences across global traditional written consent. Verbal consent, approved and populations. For instance, in a study of 6-mo-old Malawian required by the overarching and site-specific IRB boards, was also and Finnish infants, Bifidobacterium was the most common obtained by team members fluent in the local language. Samples genus despite other distinct population differences such as were collected from 11 populations (cohorts), including Ethiopia a higher relative abundance of Bacteroidetes-Prevotella and (a rural population denoted ETR; and an urban population Clostridium histolyticum in the Malawian infants (23). Murphy denoted ETU); Kenya (KE), Ghana (GN), The Gambia (a rural et al. (12) found that Staphylococcus, Escherichia-Shigella, population denoted GBR; and an urban population denoted and Veillonella were the most abundant microbial genera GBU), Peru (PE), Spain (SP), Sweden (SW), and the United States in infant feces in Ireland. This contrasts with some earlier (including a self-identified Hispanic population recruited from work. For example, Backhëd et al. (20) reported that at 4 California and denoted USC; and an ethnically heterogenous mo, Bifidobacterium, Lactobacillus, Collinsella, Granulicatella, population living in Washington/Idaho, primarily composed of and Veillonella dominated microbial communities in feces of women of northern European descent, denoted USW). Details vaginally delivered infants in Sweden, while Bifidobacterium, about these populations have been published previously (25, 26), Ruminococcus, and Bacteroidetes dominated feces of infants born and subject/sample disposition is summarized in Figure 1. A via cesarean section in the same location. (24) also reported a total of 413 mothers and their infants were enrolled. high relative abundance of Bifidobacterium in feces of Gambian For inclusion, women had to be breastfeeding or pumping ≥ infants over the first 6 mo of life, followed by Streptococcus 5 times/d and be ≥ 18 y of age. Our goal was to enroll women and members of the Enterobacteriaceae family. The common between 1 and 3 mo (± 7 d) postpartum, although 17 women taxa in many of these results suggest that there may be some were outside this target, resulting in a range from 20 to 161 d shared patterns in the bacterial community in infant feces (e.g., postpartum. Exclusion criteria for women included (1) current Bifidobacterium and Bacteroidetes as the most abundant genera indication of a breast infection or breast pain that the woman across cohorts). However, it remains unclear whether these did not consider normal for lactation, (2) illness (including fever, commonalities and differences are genuine or are simply due vomiting, severe cough, or diarrhea) in the last 7 d, and/or (3) to methodological differences. True differences, as opposed to antibiotic use in the previous 30 d. Women did not need to be biases introduced by varying methodology, might indicate that exclusively breastfeeding to participate. To be included, infants microbial communities are shaped (at least in part) by some had to be described as healthy by their mothers, have no signs combination of genetics, environment, and behavior, and that and/or symptoms of acute illness (fever, vomiting, severe cough, there may not be a universal or “normal” infant fecal microbiome diarrhea, or rapid breathing) in the previous 7 d, and have not representative of health or disease. taken antibiotics in the previous 30 d. The primary purpose of the study described here was to, using standardized collection and analysis procedures, characterize and compare human milk and infant fecal microbiomes in selected Anthropometric, Demographic, and global regions. Our hypotheses were that: (1) human milk Anthropologic Information and infant fecal microbiomes vary among cohorts representing Women’s height and weight were measured, and body mass populations from selected geographical regions; (2) there exists index (BMI) calculated. Infants were weighed, and their length a “core” group of bacteria common to milk across cohorts; measured. Infant weight-for-length z-scores were calculated (3) there exists a “core” group of bacteria common to infant using the restricted analysis function in the World Health feces across cohorts; (4) variation in the milk microbiome is Organization’s Anthro igrowup macro (27) using R (version related to variation in the infant fecal microbiome; and (5) 3.4.1). Weight-for-length z-scores flagged as biologically milk and fecal microbiomes of mothers and their own infants implausible (< −5 or > 5) were removed from the analysis (n are more similar to each other than to maternal/infant dyads = 3). Extensive in-person survey data were collected on aspects in other cohorts. Relationships of milk and fecal microbiomes of delivery, maternal, and infant characteristics; household with other important factors such as delivery mode, other composition; maternal and infant diet; and other lifestyle milk components, maternal and infant diets, and household variables. For this study, “exclusively breastfed” was defined as composition and childcare parameters will be addressed in never having received liquids (e.g., water, formula) or semi-solid subsequent publications. or solid foods. If an infant received oral medications or non- nutritive dietary supplements (e.g., gripe water, vitamin drops) MATERIALS AND METHODS at any point in his/her lifetime, but was not fed other liquids or foods, he/she was still considered exclusively breastfed. Selected Aim, Design, and Setting anthropometric and demographic data of the women and infants All study procedures were approved by the Washington for whom milk and/or fecal samples were successfully collected State University Institutional Review Board (#13264) and at and characterized are provided in Table 1.

Frontiers in Nutrition | www.frontiersin.org 3 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

FIGURE 1 | Flowchart depicting the disposition of infant feces and milk included in this study. ETR, rural Ethiopia; ETU, urban Ethiopia; GBR, rural Gambia; GBU, urban Gambia; GN, Ghana; KE, Kenya; SP, Spain; SW, Sweden; PE, Peru; USC, California (United States); USW, Washington (United States).

Infant Fecal Sample Collection If < 0.2 g of sample was available, 0.5 mL TE50 was added Fecal samples (∼1 g) were collected from 406 infants. When to the collection tube, the mixture vortexed, and the entire possible, fecal samples were collected by study personnel at the volume transferred to a sterile tube and frozen at −80◦C until same time the milk was collected; when this was not possible, DNA extraction. Frozen, homogenized fecal samples were quick- mothers collected the next fecal sample available. Samples were thawed on a dry heat block at 37◦C and vortexed to re- collected from provided diapers (Parent’s Choice; Walmart, homogenize. DNA was extracted using the QIAampR Fast DNA Bentonville, AR) or directly from the infant’s skin using a Stool Mini Kit (Qiagen, Germantown, MD) with an additional sterile scoop (Sarstedt AG & Co., Nümbrecht, Germany); the bead beating step at the beginning using 0.1mm diameter sample was then placed in the sterile polypropylene container zirconia/silica beads (BioSpec Products, Inc., Bartlesville, OK) accompanying the scoop and frozen within 30 min of collection and a FastPrep FP120A-115 (Qbiogene, Carlsbad, CA). For all (except ETR) at −20◦C. In ETR, because of unreliable electricity, rounds of DNA extractions, 500 µL TE50 taken from the same RNAlaterR (Ambion) was added to each fecal sample in a ∼1:4 aliquot used previously to prepare the samples and 500 µL ratio (feces: preservative) and frozen within 6 d. All samples were nuclease-free water (Ambion, Waltham, MA) were extracted as shipped on dry ice to the University of Idaho, Moscow, ID, USA, negative controls. Samples were eluted in 200 µL ATE buffer where they were immediately frozen at −20◦C. supplied in the kit and stored at −80◦C until amplified.

Extraction of DNA From Feces Milk Sample Collection After thawing feces at room temperature, 0.2 g of each sample Milk was collected from 412 of the women using methods was transferred into a sterile tube, 0.5 mL TE50 (10 mM Tris- described previously (26). Briefly, both participant and researcher HCl, 50 mM EDTA, pH 8) added, the mixture vortexed until wore nitrile gloves, and milk (∼30 mL) was expressed using an homogeneous, and then frozen at −80◦C until DNA extraction. electric pump (USW, USC, PE, and SW) and sterile collection

Frontiers in Nutrition | www.frontiersin.org 4 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study d ab b de 0.52) ab a bc 0.20 1 3 − 1 1.0 1.0 41 39 ab 1.1 ± ± ± abcd ± ± ± = = 44 ± 79 n n 57 68 28 167 5.2 0.91 1.6 1.30 to 26.3 − − ab a ab cde bcd ab abc 1.0 1 5 12 19 1 1.1 0.47 ab 1.1 abcd ± ± ± 61 = = ± ± ± 63 ± 50 n n 62 29 162 5.1 28.8 1.6 1.37 cd b abc de a b ab 1 4 0.26 1 1.0 23 23 1.0 cd ab 1.1 ± ± ± ± ± = = 54 ± ± 48 79 n n 49 30 1.03–0.02) (0.46–2.29) ( 169 5.0 1.4 0.53 − 25.5 − bd e b cde a a abc 1 3 0.20 1 1.1 1.0 37 40 a 1.0 0.05). abcd ± ± ± ± 48 ± ± = = ± 90 > 62 n n 34 71 0.69–0.11) ( 1.2 5.3 165 − 0.29 23.2 − ab a f ab cd 1 bc ab 1.0 1 3 d b 42 43 ± 1.0 0.19 1.1 ± ± ± = = 47 ± ± ± 31 51 n n 153 26 60 5.5 1.8 27.6 0.60 a a d bcde a cd 1 3 1 1.0 0.20 ± 42 42 d ab 1.0 ± ± (1.8–2.6) (1.5–2.1) (1.0–1.4) (1.1–1.8) (1.3–2.1) (1.4–1.9) ± 45 ± ± = = 36 ± 78 2.2 n n 73 25 abcd 159 6.0 1.03 23.4 bd b bcd bcd de bc ab 1 0.22 1 3 a d 1.0 32 37 1.0 1.1 ± ± ± ± ± = = 55 ± 92 34 ± n n (26–30)(52–66) (23–26) (67–79) (25–28) (54–67) (32–36) (64–77) (28–33) (40–57) (26–31) (52–71) (27–30) (61–75) 28 58 0.56–0.30) (0.65–1.42) (0.22–0.97) ( 159 5.2 0.13 1.9 − 24.4 − bd b cde abc abc bc ab 0.21 1 1 3 1.0 38 38 a 1.0 1.1 bcd ± ± ± ± 42 ± = = ± 97 ± 53 n n (25–28) (54–68) 0.41–0.39) ( 26 61 4.9 167 − 2.7 0.01 22.8 − The Gambia urban Ghana Kenya Peru Spain Sweden US California US Washington e abc a b cd bc ab 1 a 1.0 1 3 a 1.1 1.0 38 39 0.21 ± ± ± ± ± ± = = 53 97 ± 100 rural n n 26 65 162 3.1 5.2 21.2 The Gambia 0.51 d SEM (95% confidence interval). de ab de de ± d ab 0.22 1 1 a 4 1.0 32 34 1.0 1.1 abc ± ± ± ± ± 50 = = ± ± 100 87 n n 22 0.94–0.09) (0.11–0.91) ( 59 urban 159 0.52 5.4 1.4 − Ethiopia 21.8 − bd e ab ab ef a cd 1 a 3 1.0 0.20 1 a 40 40 1.0 1.1 ± ± ± ± ± = = 56 ± ± 95 100 n n 71 (23–26)(64–77) (20–23) (52–66) (25–28) (59–72) 24 (5.2–5.9) (5.1–5.8) (4.9–5.5) (4.6–5.1) (4.9–5.5) (5.7–6.4) (5.2–5.8) (5.0–5.6) (4.7–5.4) (4.6–5.7) (4.9–5.5) (2.5–3.5) (1.2–1.7) (2.6–3.7) (2.3–3.2) (1.6–2.3) 1.1 0.56–(0.11) ( 155 rural (153–157) (157–161) (160–164) (165–169) (157–161) (158–161) (151–154) (163–167) (166–171) (159–165) (165–169) 5.6 2.9 21.1 (20.1–22.2) (20.7–22.9) (20.2–22.2) (21.7–24.0) (23.3–25.7) (22.3–24.5) (26.4–29.0) (22.1–24.4) (23.9–27.1) (26.8–30.9) (25.0–27.7) 0.27 − Ethiopia ( − 0.05) as determined by one-way ANOVA and a distribution appropriate for the data. Values sharing a letter are not different from each other (P 1, 3 ≤ 7 1, 5 ) 1, 6 Description of the subjects for whom milk and/or fecal samples were successfully collected and characterized. 1, 9 1, 4 2 1, 10 1, 2 1 1, 8 Missing values: 1 in Ethiopia rural, 2 in The Gambia rural, 2 in The Gambia urban, 3 in Kenya, 5 in US California, 2 in US Washington Effect of cohort (P Missing values: 2 in TheMissing Gambia value: rural, 1 1 in in The TheMissing Gambia Gambia values: urban rural 1 in EthiopiaMissing rural, values: 1 2 in in Ghana, Kenya, 1Missing in 4 values: Kenya in 2 US in Washington Kenya,Missing 4 values: in 1 US in Washington EthiopiaMissing rural, value: 1 in 1 The in Gambia USMissing rural, Washington values: 1 1 in in US The California Gambia rural, 1 in Kenya, 2 in US Washington Sex (% male) Exclusively breastfed (%) Weight (kg) Weight-for-length z-score Age (y) Time postpartum (d) Parity Delivery mode (% vaginal) BMI (kg/m Height (cm) Infants TABLE 1 | Women Values are model-based estimates1 and represent means 2 3 4 5 6 7 8 9 10

Frontiers in Nutrition | www.frontiersin.org 5 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study kits (Medela, Baar, Switzerland), or hand-expressed into a positives, respectively. The first PCR was conducted in 96-well sterile collection container (all other sites). Except for those plates (USA Scientific, Ocala, FL), using the Veriti model thermal collected in ETR where there was unreliable electricity for cold cycler (Applied Biosystems, Foster City, CA) under the following storage, samples were immediately placed on ice, aliquoted into conditions: initial denaturation at 98◦C for 30 s; followed by 15 polypropylene cryotubes (Simport Scientific, Saint-Mathieu-de- cycles of denaturation at 98◦C for 10s, annealing at 51◦Cfor20s, Beloeil, Quebec) within 30 min, and frozen at −20◦C. Milk and extension at 72◦C for 20s. After the 15th extension cycle, collected in ETR was preserved in a 1:1 ratio with Milk the thermal cycler was paused at 72◦C and the samples removed. Preservation Solution (Norgen Biotek, Thorold, Ontario) and For the second step, 2 µL of a unique barcoded primer pair with frozen within 6 d. We have shown previously that this method Illumina adaptors attached [2 µM; obtained from the University can maintain bacterial DNA integrity in human milk held at of Idaho’s Institute for Bioinformatics and Evolutionary Studies 37◦C for at least 2 wk (28). All samples were shipped on dry ice (IBEST) Genomics Core facility] were then added to each sample. to the University of Idaho, where they were immediately frozen The plate was vortexed, briefly centrifuged, placed in the thermal at −20◦C. cycler, heated to 98◦C for 30s, amplified for an additional 15 cycles as described above (except with an annealing temperature Extraction of DNA From Milk of 60◦C), subjected to a 2-min final extension step at 72◦C, and For all sites except ETR, 1 mL milk was thawed on ice and held at 4◦C until the plate was removed from the thermal cycler. centrifuged (13,000 x g) for 10min at 4◦C. After removing the lipid and supernatant layers, the cell pellet was resuspended DNA Quality, Quantification, and Pooling in 500 µL TE50. Samples were subjected to enzymatic lysis PCR products for all negative and positive controls were by adding 100 µL of a mixture containing 50 µL lysozyme electrophoresed at 80 V on a 1% agarose gel for 30 min in tris- (10 mg/mL in nuclease-free water) (Sigma-Aldrich), 6 µL acetate-ethylenediamine tetraacetic acid (EDTA) buffer (TAE; mutanolysin (25 KU/mL in nuclease-free water) (Sigma- 40mM Tris, 20mM acetic acid, 1mM EDTA); stained with Aldrich), 3 µL lysostaphin (4,000 U/mL in 20mM sodium GelRedTM (10X, Biotium, Fremont, CA); and run alongside acetate) (Sigma-Aldrich), and 41 µL TE50 for 1h at 37◦C. a 1-kb ladder (Thermo Fisher Scientific, Grand Island, NY). Following the enzymatic lysis, samples were subjected to physical Gels were visualized using an UltraCam Digital Imaging System disruption by bead beating with 0.1 mm zirconia/silica beads (BioRad, Hercules, CA). Amplicon quality was assessed using (BioSpec Products) for 1 min on setting 5 using a FastPrep FP120 the QIAxcel DNA Screening cartridge (Qiagen). Briefly, 2 µL (Qbiogene). DNA was extracted using a modified protocol of the PCR product generated in the second amplification was added DNA Mini Kit (Qiagen), whereby 100 µL 3M sodium acetate, to 8 µL QX DNA dilution buffer and visualized using high- pH 5.5, was added to the lysate prior to addition to the spin resolution capillary electrophoresis on a QIAxcel Advanced column. DNA was eluted in 50 µL nuclease-free water (Ambion). System (Qiagen). Samples with a peak at the appropriate For each set of milk samples processed in this way, 500 µL TE50 amplicon size and minimum levels of primer dimers were was extracted as a negative control. deemed as having adequate quality. DNA (2 µL amplicon DNA in 0.25 mL of each milk sample collected in combined with 198 µL of a solution Qubit dsDNA HS reagent ETR was extracted using the kit accompanying the Milk and Qubit HS buffer in a 1:200 dilution) was quantified using Preservation Solution (Norgen Biotek, Thorold, Ontario) as per the QubitR 2.0 Fluorometer and the QubitTM dsDNA High manufacturer’s instructions, including the 2 h enzymatic lysis Sensitivity Assay (Thermo Fisher Scientific, Waltham, MA). (20 mg/mL lysozyme). DNA was eluted in 100 µL elution buffer Samples were pooled to contain 50 ng DNA from each sample. (included with the kit) and stored at −20◦C until amplification. If samples amplified poorly, an attempt at re-amplification was Nuclease-free water (500 µL; Ambion) was extracted as a made. Subsequently, amplicons were combined if necessary to negative control. obtain sufficient amounts of DNA, or the entire volume from the PCR was used. Amplification of Bacterial DNA For both milk and feces, a dual-barcoded, two-step 30-cycle Sequencing and Identification of polymerase chain reaction (PCR) was conducted to amplify the Microbial DNA V1-V3 hypervariable region of the 16S rRNA bacterial gene. Amplicon pools were size-selected using AMPure beads For the first step, a 7-fold degenerate forward primer targeting (Beckman Coulter, Indianapolis, IN); quality checked on a position 27 [modified from Frank et al. (29)] and a reverse Fragment Analyzer (Advanced Analytical Technologies, Inc., primer targeting position 534 (positions numbered according Ankeny, IA); and quantified using the KAPA Biosciences to the Escherichia coli rRNA gene) were used as described Illumina library quantification kit and Applied Biosystems previously (30). The reaction mixture included 12.5 µL Q5R StepOne Plus real-time PCR system. Pools of PCR amplicons Hot Start High-Fidelity 2X Master Mix (New England BiolabsR , for milk and feces were sequenced in two separate sequencing Inc., Ipswich, MA); 0.25 µL each forward and reverse primers runs by sample type. Sequences were obtained using an Illumina (10 µM each); 2 µL template DNA; and 8 µL nuclease-free, MiSeq (San Diego, CA) v3 paired-end 300-bp protocol for 600 sterile water (Ambion) to bring the reaction volume for the cycles at the IBEST Genomics Core. first PCR to 23 µL. Nuclease-free water (2 µL) and Escherichia Sequence reads were demultiplexed using dbcAmplicons coli DNA (2 µL; 221 ng/mL) were used as PCR negatives and (a custom python application; https://github.com/msettles/

Frontiers in Nutrition | www.frontiersin.org 6 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study dbcAmplicons). During preprocessing, barcodes were allowed ≤ indices (specifically, richness) are linked to the number of reads 1 mismatch (hamming distance), and primers were allowed ≤ 4 per sample, we rarefied the infant fecal and milk data to 1,000 mismatches (Levenshtein distance) as long as the final 4 bases reads prior to calculation of all diversity indices. of the primer perfectly matched the target sequence. Sequence reads without a corresponding barcode and primer sequence Statistical Analyses were discarded. Sequence reads were also trimmed of their All analyses using SAS software were conducted in version 9.3 primer sequence. Reads were split into separate sample R1 and (SAS Institute Inc., Cary, NC); all other analyses were performed R2 files using a custom python script (splitReadsBySample.py; in R [version 3.4.1; (42)]. Significance for all statistical tests were https://github.com/msettles/dbcAmplicons/blob/master/scripts/ declared at P ≤ 0.05. Prior to performing inferential statistical python/splitReadsBySample.py). Sequence reads were evaluated tests, all relative abundance data were rounded to the tenth for quality, trimmed, and filtered using the DADA2 sequence decimal place. Any taxa denoted as having a relative abundance process pipeline (version 1.2.2; (31)). The output from DADA2 of zero for all sites after rounding were excluded from further infers amplicon sequence variants (ASV) after modeling the analyses and were not included in tables. For the remaining errors from a subset of reads (1 × 106) from the sequencing run. taxa, all zero values in the dataset were replaced with 1 × 10−6. Each MiSeq run was analyzed separately for error estimation. One-way analysis of variance (ANOVAs) were carried out using Briefly, sequence reads were truncated to 270 bases with a maxEE a generalized linear mixed model (GLIMMIX; SAS) assuming setting of 4. Reads were also truncated if the base call reached Q2. distributions appropriate for the response types. In the case of Reads with < 270 bases were discarded. Because of loss of reads continuous descriptive variables presented in Table 1, all data and little overlap between forward and reverse reads following except for time postpartum, maternal height, and weight-for- quality filtering and trimming, only forward reads were used in length infant z-scores (which were all normally distributed) subsequent analyses. ASV were then assigned taxonomies using were assumed to be log-normally distributed; for binary the Ribosomal Database Project (RDP) Bayesian classifier (32) variables, we assumed a binomial distribution, and for relative and the SILVA 16S rRNA database version 123 formatted for abundance (proportional data) we utilized a beta distribution. DADA2 (33–35). Relative abundances of bacterial taxa at various For infant fecal diversity indices, data were assumed a log-normal taxonomic levels were calculated from sequence read count data transformation; for milk, richness, inverse Simpson, and Fisher in R (version 3.4.1) using the phyloseq package (36). diversity values were assumed a log-normal transformation; Shannon diversity values were untransformed. For all ANOVA Designation of Core, Unique, and and associated pairwise comparisons, a Bonferroni correction Aggregated Taxa for multiple comparisons was applied. In the case of multiple A set of “core” genera were characterized for each sample type comparisons, P-values presented are Bonferroni’s adjusted P. both in the overall dataset and within each cohort. To be included Hierarchical clustering was performed on relative abundance in the core taxa, a genus must have been present in ≥ 90% of data using the vegan package (43), specifically the vegdist the samples and represent ≥ 0.1% of all identified taxa. We also function and hclust function, in R using a Bray-Curtis characterized the relatively “unique” taxa across all taxa identified dissimilarity matrix and average linkage hierarchical clustering. in the sequencing run for each sample type, defined as genera Non-metric multidimensional scaling (NMDS) plots were present in only one cohort and in ≥ 10% of samples within created in R using the vegan and ggplot2 (44) packages that cohort. using rounded data and the aggregate taxa lists for milk and The 10 most-abundant genera from each cohort and for each infant feces. sample type (based on relative prevalence) were identified and For analyses exploring associations between microbial combined to create a set of 28 genera for infant fecal data and a communities of milk and infant fecal samples, only matched- set of 29genera for the milk data. An “other”category was created dyad samples were included (n = 360). Heatmaps based upon at both the phylum and genus levels in these datasets, which Spearman rank correlations were constructed in R using the is a sum of all other identified taxa within each dataset. These stats (42) and gplots (45) packages to evaluate relationships datasets are hereafter referred to as “aggregations” in subsequent between the 28 aggregated genera for infant fecal samples and text and figures. the 29 aggregated genera in milk. Canonical correlation analysis was performed in SAS to explore communities in the data set Calculation of Diversity Indices as a whole (all cohorts combined) using the aggregated infant Diversity indices (richness, Shannon diversity, inverse Simpson fecal and aggregated milk genera. For canonical correlations, diversity, and Fisher diversity) were calculated using phyloseq relative abundance data within each observation were first (36) and sequence read count data. Richness measures the transformed using a logit transformation. To compare the absolute number of taxa present in a population (37), whereas, within- and between-group similarity of bacterial communities, Shannon diversity is a compound measure of richness and an analysis of similarity (ANOSIM) was performed in R evenness (38, 39). Inverse Simpson diversity is the inverse of the with the vegan package using Bray-Curtis distance and 999 probability that two randomly chosen taxa belong to different permutations. To evaluate maternal/infant dyad similarity, Bray- genera (38, 40), and Fisher diversity describes the mathematical Curtis dissimilarity and Jaccard indices were calculated for all relationship between the number of genera and the number matched dyads and all combinations of non-matched dyads, and of individuals within each genus (41). Because some diversity Wilcoxon test was used to determine differences between these

Frontiers in Nutrition | www.frontiersin.org 7 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study values (14). Associations between diversity indices in milk and filtering of any read that could not be classified to the genus infant feces were evaluated using Spearman rank correlations. level, and omitting any sample with < 1,000 reads, the infant fecal dataset analyzed here contained 4,314,551 reads across 377 RESULTS samples, with a mean (± SD) of 11,444 ± 6,198; and a range of 1,662–40,255 reads. For the 409 milk samples, sequencing Description of Subjects generated 7,528,193 reads, with a mean (± SD) of 18,406 ± There were many notable differences among the cohorts, as 18,389 reads and a range of 12−141,620 reads following initial summarized in Table 1. For example, Spanish women were older processing using the DADA2 workflow. Using the same filtering than women in ETR, ETU, GBR, GBU, GN, KE, PE, and USW criteria as the infant fecal dataset, the milk dataset used here (P ≤ 0.0017); women in SW were breastfeeding younger infants contained 6,709,277 reads across 394 samples, with mean (± SD) than women in ETR, KE, SP, and USW (P ≤ 0.0328); and there of 17,029 ± 16,783, and a range of 1,302–130,700 reads. These were fewer vaginally-delivered infants in PE than in ETR, ETU, curated datasets were used for all further analyses. GBR, GBU, GN, and SP (P < 0.0001). Women in GBR and ETR had a higher parity than women in all cohorts except GBU and KE (P = 0.0411). Maternal BMI was higher in women from PE Infant Fecal Microbiome: Individual and USC than women from all African cohorts and SP (P ≤ Phyla Analysis 0.0113), and women in SW were taller than women in ETR, ETU, Pie charts illustrating the mean relative abundances of bacterial GBR, GN, KE, PE, and USC (P ≤ 0.0174). In addition, exclusive phyla in infant feces are provided in Figure 2A (mean values breastfeeding was more common in ETR, ETU, and GBR than available in Supplementary Table 2). Overall, Firmicutes, in GN, KE, and PE (P ≤ 0.0052). Kenyan infants were heavier Proteobacteria, Bacteroidetes, and Actinobacteria together than infants in GBR, GBU, GN, SP, SW, and USW (P ≤ 0.0231); composed > 99.5% of the bacteria identified, though notably accordingly, infants in KE had a higher weight-for-length z-score only 5 phyla were identified in infant feces; the “other” category, than infants in ETR, ETU, GBU, GN, SP, SW, and USW (P ≤ defined as all identified phyla besides the top 4 described above, 0.0157), though notably, the average z-score for all cohorts was was entirely composed of Verrucomicrobia (∼0.5%). To analyze within the normal range (-2 < z < 2). differences in the relative abundances of these phyla among cohorts, ANOVA was performed and indicated an effect of Sequencing Summary cohort existed for Firmicutes and Actinobacteria. The relative The sequencing run for the 398 infant fecal samples generated abundance of Firmicutes was higher in feces from ETR than 4,385,982 reads, with a mean (± standard deviation, SD) of feces from GN, SP, and USW (P ≤ 0.0374) and Actinobacteria 11,020 ± 6,632 reads and a range of 10 to 40,267 reads following was lower in feces from ETR, KE, PE, and USC than feces from initial processing using the DADA2 workflow. After additional GN (P ≤ 0.0360).

FIGURE 2 | Mean relative abundances of the bacterial (A) phyla and (B) an aggregation of the 10 most-abundant bacterial genera in each cohort in infant feces. ETR, rural Ethiopia; ETU, urban Ethiopia; GBR, rural Gambia; GBU, urban Gambia; GN, Ghana; KE, Kenya; SP, Spain; SW, Sweden; PE, Peru; USC, California (United States); USW, Washington (United States).

Frontiers in Nutrition | www.frontiersin.org 8 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

Infant Fecal Microbiome: Individual GN, and KE, and Bacteroides was part of the core taxa for Genera Analysis USW. There were no unique bacterial genera identified within There was also variation in the relative abundance of any cohort. bacterial genera both among (Figure 2B; Table 2) and within (Supplementary Figures 1–11) cohorts. There was a statistical Infant Fecal Microbiome: effect of cohort on 8 of the 28 aggregate genera, 7 of which were Diversity Measures among the most abundant taxa (Table 2). Differences between Microbial diversity of infant feces also varied by cohort (Table 4). cohorts varied by genera. For example, feces from GN, SP, SW, Richness and Fisher diversity of KE feces were higher than those USC, and USW had lower relative abundance of Streptococcus of feces from ETU, GN, PE, SP, and USC (P ≤ 0.0127 and ≤ than feces from GBR and GBU (P ≤ 0.0012). Feces from infants 0.0111, respectively). Conversely, richness and Fisher’s diversity in USW had lower relative abundance of Escherichia/Shigella score of fecal samples collected in USC were the lowest of all than those in ETU and PE (P ≤ 0.0399). Feces from infants in cohorts (P ≤ 0.0126, P ≤ 0.0164, respectively). This finding, GN had a lower relative abundance of Veillonella than infants in however, may be due to the small sample size in the USC ETR, GBR, KE, PE, and SP (P ≤ 0.0105). Conversely, GN infants’ cohort and should therefore be interpreted cautiously. Feces from feces contained relatively more Bifidobacterium than feces from infants in KE had higher Shannon diversity than those collected ETR, PE, and USC infants (P ≤ 0.0310). Relative abundance of inGNandUSC(P ≤ 0.0448). Feces from GN had a lower inverse Bacteroides was higher in feces from infants in USW than GBR, Shannon diversity score than those from ETR, ETU, GBR, GBU, GBU, and KE (P ≤ 0.0403). Feces from infants in the two rural GN, KE, and PE (P ≤ 0.0354). populations (ETR and GBR) had higher relative abundance of Lactobacillus than feces from PE, SP, SW, USC, and USW (P ≤ Milk Microbiome: Individual Phyla Analysis 0.0122). Relative abundance of Clostridium sensu stricto 1 was A total of 15 phyla were identified in milk, with Firmicutes, higher in SW than GBU (P = 0.0350), and feces from infants in Proteobacteria, Actinobacteria, and Bacteroidetes collectively GN had a higher relative abundance of Enterococcus than feces representing 97.7% of those identified (Figure 4A; from infants in ETR, ETU, GBR, KE, PE, USC, and USW (P ≤ Supplementary Table 3). There was an effect of cohort on 0.0347). For more details, see Table 2. Firmicutes, Proteobacteria, Actinobacteria, Bacteroidetes, and “other.” ANOVA analysis indicated that the relative abundance Infant Fecal Microbiome: of Firmicutes was lower in ETR than in all cohorts besides GBR and USC (P ≤ 0.0139). Proteobacteria was relatively more Community Analysis abundant in milk collected in ETR than all other cohorts (P ≤ Considering similarities and differences among the collective 0.0077). Actinobacteria in milk was more abundant in ETR, relative abundances of the aggregate genera, hierarchical ETU, GBR, and GBU than in GN, USC, SP, and PE (P ≤ 0.0014). clustering patterns (Figure 3A) suggest that fecal bacterial Relative abundance of Bacteroidetes was higher in KE than GN community structure in USC and SW, as compared to African (P = 0.0060). There was also an effect of cohort on the “other” cohorts, clustered together and were characterized by relatively category; relative abundance in GBR was higher than in ETR, high amounts of Bacteroides, Clostridium sensu stricto 1, and ETU, GN, PE, SP, SW, and USC (P ≤ 0.0036). Parabacteroides, but relatively low amounts of Lactobacillus. Fecal microbial communities in GBU were closely related to Milk Microbiome: Individual those in GBR and GN, a clade characterized by high relative Genera Analysis abundance of Streptococcus, Bifidobacterium, and Lactobacillus, There was also variation in milk microbiome at the genus level and low abundance of Bacteroides. GN differed from GBU and both within and among cohorts (Supplementary Figures 14–24 GBR in that Veillonella was low in GN. and Figure 4B, respectively). Of the 29 genera evaluated, NMDS plots suggested no clear clustering by cohort ANOVA indicated that there was an effect of cohort on 19 of (Supplementary Figure 12A). Although there was considerable them (Table 5). Examples of differences among cohorts include variation in fecal bacterial composition among infants within a a higher relative abundance of Rhizobium, Achromobacter, and cohort (Supplementary Figures 1–11), there was more similarity Psychrobacter in milk collected in ETR than all other cohorts (P < within a cohort than among a random subsampling across 0.0001). Except for ETU, milk collected in ETR also had a higher = < cohorts (ANOSIM R 0.1318, P 0.001). In other words, relative abundance of Corynebacterium 1 than all other cohorts (P samples within a cohort were more similar to each other than ≤ 0.0417). Peruvian milk bacterial communities, on average, were would be expected by random chance. comprised of 50% Streptococcus, which was relatively higher than all African (ETR, ETU, GBR, GBU, GN, and KE) and US (USW Infant Fecal “Core” and “Unique” Bacteria and USC) samples (P ≤ 0.0386). Milk from women in USW had Overall, Streptococcus, Escherichia/Shigella, and Veillonella relatively more Dyella than all sites except The Gambia (GBR and were identified as core taxa in infant feces, being present in GBU) (P ≤ 0.0040). 98.4, 91.7, and 90.2% (respectively) of all samples (Table 3; Supplementary Figure 13). When considered within each Milk Microbiome: Community Analysis cohort, there were sometimes different sets of core taxa. For Hierarchical clustering of the complex bacterial community example, Lactobacillus was part of the core taxa for ETR, GBR, structures in milk (Figure 3B) suggested that samples

Frontiers in Nutrition | www.frontiersin.org 9 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study ton c b ab ab a b ab b 41 1.8 1.6 1.6 1.5 2.8 1.0 0.7 0.4 0.4 0.8 0.4 0.2 0.1 0.2 0.2 0.1 0.1 0.3 0.1 0.2 0.1 0.1 0.1 0.1 0.1 0.0 0.0 0.0 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 9.6 5.6 4.1 1.4 12.1 10.0 10.2 18.1 c ab ab b ab bc ab b 12 2.0 2.6 3.3 1.7 3.8 1.3 1.8 0.7 3.8 0.9 1.9 0.7 2.1 0.3 0.5 1.8 0.3 1.0 0.2 0.8 0.3 0.8 0.2 0.8 0.1 0.4 0.1 0.6 0.1 0.5 0.2 0.9 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.4 0.2 0.5 0.1 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 7.1 9.2 5.6 4.5 5.9 1.0 12.4 13.4 c ab ab ab ab b a ab 23 2.4 1.8 2.2 3.2 2.0 1.0 1.4 0.7 2.1 0.6 2.8 0.9 2.4 0.3 0.4 1.8 0.3 0.9 0.2 0.8 0.2 0.8 0.2 0.6 0.1 0.4 0.1 0.5 0.1 0.5 0.2 0.7 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 9.0 9.5 4.6 6.1 1.3 12.0 10.9 15.2 c ab a ab ab b ab ab 37 1.9 2.6 2.1 2.0 2.1 0.9 0.6 2.8 0.4 2.7 0.5 3.6 0.7 0.4 1.6 0.2 1.0 0.1 0.9 0.2 0.8 0.1 0.7 0.3 0.1 0.4 0.1 0.5 0.1 0.5 0.2 0.7 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.4 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 5.1 3.7 1.5 12.5 16.7 13.6 11.7 12.0 bc a a b ab b ab b 42 2.1 2.7 2.1 1.8 1.5 1.0 0.6 0.6 2.7 0.6 2.2 0.4 2.4 0.2 0.3 1.7 0.2 0.9 0.2 0.7 0.2 0.8 0.2 0.6 0.1 0.6 0.1 0.5 0.1 0.6 0.2 0.7 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 9.3 5.8 3.0 1.3 15.7 19.9 14.3 10.8 bc ab a ab b ac ab b 42 2.2 2.7 2.0 1.6 1.5 1.9 0.5 0.6 3.3 0.5 3.3 0.4 2.0 0.2 0.2 1.3 0.3 1.1 0.1 0.8 0.2 0.9 0.1 0.8 0.1 0.5 0.1 0.5 0.1 0.5 0.2 0.7 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.4 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 9.9 8.8 2.8 1.1 16.5 19.0 14.3 12.5 0.05). c ab b a ab ac ab a > 32 2.0 2.8 3.1 1.0 1.8 2.2 0.6 0.6 0.5 2.9 0.6 2.5 0.5 2.0 0.2 1.2 0.2 1.5 0.1 0.6 0.2 1.1 0.1 0.6 0.1 0.6 0.1 0.5 0.1 0.5 0.2 0.8 0.1 0.4 0.0 0.3 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 5.6 9.6 3.0 3.1 12.3 17.5 19.9 12.8 a ab ab ab b ab b ab a urban Ghana Kenya Peru Spain Sweden US California US Washing 38 3.4 2.1 1.9 2.5 1.3 1.6 0.5 0.3 0.5 2.3 0.7 2.7 0.4 2.1 0.2 1.1 0.3 0.9 0.2 0.7 0.2 0.8 0.1 0.7 0.2 0.4 0.1 0.5 0.1 0.5 0.2 0.8 0.1 0.4 0.1 0.2 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n different from each other (P 7.3 9.3 2.5 1.5 33.6 13.6 11.9 16.0 SEM. ab ab a ab b a ab b ± 28 bacterial genera in infant feces. 38 3.1 2.1 2.1 2.4 1.4 2.3 0.6 0.2 0.4 2.4 0.5 3.4 0.4 2.1 0.2 1.1 0.2 1.3 0.2 1.1 0.2 0.9 0.1 0.6 0.2 0.8 0.1 0.5 0.1 0.5 0.2 0.8 0.1 0.4 0.1 0.5 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 s = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 7.7 2.8 1.2 27.7 13.5 13.2 15.1 14.7 bc a ab ab ab ab ab b 32 2.4 3.3 1.9 1.7 1.8 1.9 0.6 0.6 2.2 0.5 2.3 0.5 2.1 0.3 0.2 1.1 0.2 1.0 0.1 0.9 0.2 0.9 0.2 0.6 0.1 0.8 0.1 0.5 0.1 0.5 0.2 0.8 0.1 0.4 0.1 0.5 0.1 0.4 0.1 0.3 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 9.8 9.4 3.0 1.2 16.0 22.0 10.5 10.6 bc ab a b ab a ab b 40 2.4 2.7 2.0 1.4 1.7 2.3 0.6 0.5 2.7 0.4 2.6 0.4 2.3 0.3 0.2 1.1 0.2 1.0 0.1 0.7 0.2 0.9 0.1 0.7 0.1 0.5 0.1 0.5 0.1 0.6 0.2 0.8 0.1 0.4 0.1 0.3 0.1 0.4 0.1 0.4 0.1 0.5 0.0 0.2 0.0 0.2 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 8.7 9.6 3.2 2.4 1.8 2.1 1.4 1.1 1.0 0.7 0.8 0.6 0.4 0.5 0.5 0.8 0.5 0.3 0.4 0.3 0.5 0.2 0.2 0.1 18.2 19.0 13.7 15.2 0.05) as determined by one-way ANOVA; values sharing a letter are not 1 ≤ 1 1 1 1 1 Effect of cohort on relative abundances (%) of the aggregate 1 1 Effect of cohort (P TABLE 2 | GenusStreptococcus Ethiopia rural Ethiopia urban The Gambia rural The Gambi Values are model estimates based on a beta distribution and represent mean 1 Escherichia/Shigella Veillonella Bifidobacterium Bacteroides Lactobacillus Clostridium sensu stricto 1 Lachnoclostridium Klebsiella Parabacteroides Enterococcus Enterobacter Megasphaera Staphylococcus Prevotella 9 Erysipelatoclostridium Rothia Citrobacter Akkermansia Campylobacter Lactococcus Gemella Tyzzerella 4 Blautia Paucisalibacillus Leuconostoc Acinetobacter Acetobacter

Frontiers in Nutrition | www.frontiersin.org 10 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

FIGURE 3 | Hierarchical clustering (vertical axis) of the mean relative abundances for an aggregation of the 10 most-abundant bacterial genera from each cohort in (A) infant feces and (B) milk. ETR, rural Ethiopia; ETU, urban Ethiopia; GBR, rural Gambia; GBU, urban Gambia; GN, Ghana; KE, Kenya; SP, Spain; SW, Sweden; PE, Peru; USC, California (United States); USW, Washington (United States). collected in GBR and GBU milk clustered together with several unique bacteria identified in some cohorts in milk: only high abundance of Streptococcus, and intermediate abundances milk collected in ETR contained Acidothermus, Demequina, of Staphylococcus, Corynebacterium 1, Dyella, and Bacillus. PE Flaviflexus, and Pediococcus; milk from GBU uniquely contained and SP were relatively similar, characterized by high relative Chroococcidiopsis and Isoptericola; and milk from GN uniquely abundance of Streptococcus, Staphylococcus, and Rothia, and contained Akkermansia and Butyricicoccus. low abundance of Corynebacterium 1. It is noteworthy that ETR formed an outgroup characterized by relatively high Milk Microbiome: Diversity Measures abundance of Rhizobium and Achromobacter; intermediate There was an effect of cohort on all indices of milk microbial abundances of Streptococcus and Staphylococcus; and very little diversity considered (P < 0.0001; Table 4). Milk collected in ETR Propionibacterium, Dyella, and Rothia. Samples from GN also had a higher richness and a higher Fisher diversity scores than formed a unique clade with relatively high levels of Lactobacillus, milk from ETU, GN, PE, SP,SW, USC, and USW (P ≤ 0.0181; P ≤ Klebsiella, and Enterococcus. 0.0106, respectively). The mean Shannon diversity score of milk Like feces, there was more similarity in the milk bacterial from ETR was higher than those of milk from GN, PE, SP, SW, composition within a cohort than across cohorts (ANOSIM R and USC (P ≤ 0.0182). Inverse Simpson diversity scores followed = 0.2244, P = 0.001). NMDS plots were created to evaluate similar trends as Shannon diversity. the similarity of cohorts, but no clear clustering was observed (Supplementary Figure 12B). Relationships Between Milk and Fecal Microbiomes Milk Microbiome: Core and A few simple relationships (P ≤ 0.01, −0.3 ≥ rs ≥ 0.3) were Unique Bacteria observed between relative abundances of the 28 aggregated For milk, Staphylococcus and Streptococcus were identified as genera in infant fecal and 29 aggregated genera in milk core genera, being present in 98.7 and 97.7% of all samples, samples (Figure 5). Relative abundances of Psychrobacter respectively (Table 3; Supplementary Figure 13). However, as and Achromobacter in milk were positively correlated with with feces, there were sometimes different sets of core taxa within Leuconostoc in feces, and the relative abundance of Lactobacillus each cohort: Propionibacterium was also present in 90.5% of in milk positively correlated with Lactobacillus in feces. milk from KE, Dyella in 94.9% of milk collected in USW, and On a multivariate basis, canonical correlations between the Corynebacterium 1 in 95.0 and 94.1% of milk samples collected aggregated milk genera and the aggregated infant fecal genera in ETR and ETU, respectively. In addition, it is noteworthy support a strong relationship between these varied and complex that milk collected in ETR contained 8 core genera, including communities (r1 = 0.663 P < 0.0001; Figure 6). In this analysis, Rhizobium, Brevundimonas, and Achromobacter, which were the first canonical component for infant feces were largely driven present in every sample from this location. There were also by Lactobacillus (rs = 0.508) and Leuconostoc (rs = 0.537), and

Frontiers in Nutrition | www.frontiersin.org 11 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study ab abc abc abc abc bcd 377 394 bc abc 1 = = 91.7 90.2 66.0 67.4 41 39 1 1.09 1.04 0.13 1.06 1.13 1.13 N N ± = = ± ± US ± ± ± ± ± n n 44 34 erall datasets Washington 7.25 2.37 2.67 6.83 7.95 9.47 ton Overall d bc ab cd de abc 41 39 d cd = = 97.6 98.4 87.8 95.1 94.9 1 1 1.12 12 19 1.07 1.17 0.18 1.19 n n 1.19 ributed, were back ± ± ± = = ± ± ± ± ± n n 20 28 California 3.65 1.95 5.67 2.33 5.30 6.22 ab cd bc abc abc cde bc abc 12 19 1 23 23 1 1.12 0.16 1.17 1.05 1.08 1.16 = = ± 75.0 80.5 83.3 91.7 = = ± ± ± ± ± ± ± n n n n 36 35 8.22 2.34 5.46 2.51 7.02 7.40 c abc ab cd abc bcd 23 23 c bc 1 1 = = 1.06 37 40 91.3 87.0 1.09 1.12 1.04 0.13 1.13a ± n n ± ± = = ± ± ± ± ± n n 30 39 5.78 7.18 8.07 2.36 2.52 7.08 37 40 = = 86.5 a cd bc n n abc bcd bcd bc bc 90% of the samples in a single cohort (cohort core) or in the ov 1 1 1.09 42 43 0.12 1.06 1.04 1.12 1.12 ≥ ± ± ± = = 42 43 ± ± ± ± ± n n n = = 32 42 64.3 48.6 47.8 41.7 61.0 76.7 79.1 82.5 87.5 78.9 84.6 75.5 n n 7.82 2.50 6.35 2.48 6.49 8.87 es and, except for Shannon diversity which was normally dist a ab a a 42 42 abc abc a ab = = 100 100 100 100 100 97.4 97.7 1 1.09 90.5 92.9 94.6 85.7 85.7 89.2 82.6 50.0 87.8 87.0 90.5 1 1.12 1.04 1.06 42 42 n n 0.12 1.12 ± ± ± ± ± ± = = ± ± n n 42 59 0.05). 32 37 2.76 8.96 10.14 2.96 9.36 > 14.00 = = 96.9 95.2 97.6 75.0 90.6 92.9 81.6 n n c e b b d bc d bc 1 1 1.13 1.13 1.04 1.10 0.13 32 37 1.07 ± ± ± ± ± ± ± = = ± n n Ghana Kenya Peru Spain Sweden US 21 32 5.44 3.93 2.08 5.04 2.05 a urban Ghana Kenya Peru Spain Sweden US California US Washing 6.42 38 38 = = 86.8 a a abc n n ac ab ab ab ab 1.13 1 1 1.09 38 38 1.13 1.04 0.13 1.06 different from each other (P ± ± ± ± = = ± ± ± ± n n urban 38 56 8.38 2.60 3.11 7.97 12.08 The Gambia 12.98 ad a abc ac ab abc unts of niche-specific “core” genera, defined as being present i 38 39 ab abc = = 89.5 86.8 94.7 94.7 92.1 97.4 100 97.4 82.1 84.2 21.1 83.3 81.4 52.5 75.0 68.4 71.8 73.7 1 1.13 1.09 38 39 1 1.13 1.04 0.13 000 reads) of infant feces and milk. Values are model estimat 1.06 n n ± ± ± = = ± ± ± ± ± ” bacteria. n n rural 57 37 8.88 2.62 3.09 7.52 11.44 The Gambia 13.47 bc a ac bc abc abc b bc 1 1 1.14 1.10 32 34 1.04 32 34 1.07 0.14 1.14 ± ± ± ± = = ± = = ± 2.9 5.1 2.6 2.6 0 2.3 0 0 5.3 5.1 12.4 ± ± 84.4 87.5 11.8 2.6 13.2 2.6 16.7 7.0 15.0 8.3 15.8 15.4 19.7 11.8 30.8 28.9 31.1 4.8 14.0 25.0 4.2 21.1 20.5 25.5 61.8 56.4 47.4 7.9 73.8 44.2 55.0 29.2 36.8 30.8 50.3 76.5 43.6 65.8 10.5 69.0 18.6 17.5 12.5 0 12.8 40.7 n n n n 52 31 urban Ethiopia 8.64 2.55 6.20 2.90 9.70 11.84 a ab a a abc abc a abc 1 1.12 1.13 0.13 1.09 40 40 1 1.04 1.06 ± ± ± 40 40 ± ± = = ± ± ± n n 7.5 64.7 84.6 84.2 34.2 83.3 86.0 80.0 62.5 42.1 = = 100 96.9 100 96.9 100 100 100 100 100 97.3 95.7 100100 100 100 92.3 100 97.4 97.4 100 100 100 100 100 100 100 98.7 100 100 100 85 95.0 80.0 81.3 92.5 70.0 53.1 47.4 42.1 75.0 54.8 69.0 70.3 87.0 75.0 75.0 82.4 74.4 68.4 31.6 95.0 94.1 92.5 92.5 rural 36 n n 3.16 7.94 22.71 2.51 7.42 10.32 SEM. ± 0.05) as determined by one-way ANOVA; values sharing a letter are not ≤ 1 1 1 1 Effect of cohort on microbial diversity indices (rarefied to 1 Percent of infant feces and milk samples with quantifiable amo 1 1 1 1 Effect of cohort (P Richness Shannon diversity Richness (overall core). Shannon diversity Inverse Simpson diversity Veillonella TABLE 3 | GenusInfant Feces Ethiopia rural Ethiopia urban The Gambia rural The Gambi Bolded values are genera considered part of the cohort or overall “core TABLE 4 | IndexInfant Feces Ethiopia Values represent means transformed from estimated log values. Escherichia/Shigella Fisher diversity Streptococcus 1 Bifidobacterium Inverse Simpson diversity Milk Lactobacillus Fisher diversity Bacteroides Milk Staphylococcus Streptococcus Propionibacterium Achromobacter Corynebacterium 1 Acinetobacter Dyella Brevundimonas Kocuria Rhizobium

Frontiers in Nutrition | www.frontiersin.org 12 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study the first canonical component in milk was driven by Lactobacillus Based on the summary statistics calculated from (rs = −0.498). When plotted by cohort, the relationships among anthropometric and health data evaluated, we believe our the infant fecal and milk taxa appear to be specific to cohort; for data are representative of generally healthy women and their example, the relationship (using the first canonical component) infants in the locations studied. However, due to the high level between milk and infant fecal microbiomes was strong in PE but of individual microbial variation present among these samples, diminished in SW. Correlations between milk and infant fecal even within a cohort, these data may not be representative of diversity indices did not reveal significant correlations in any neighboring populations or countries. Indeed, Meehan et al. (16) cohort except for KE where richness and Fisher diversity of infant determined that among foragers and horticulturalist women feces and milk were positively correlated with each other (r = in the Central African Republic who spend considerable time 0.31, P = 0.0453; Supplementary Table 4). in in proximity to each other, milk microbial communities vary significantly both within populations and between ethnic Microbial Relationships Between Women groups. In the present study, even though Kenya and Ethiopia and Their Infants are geographical neighbors, substantial differences in microbial communities existed. For example, Veillonella and Lactobacillus Upon evaluating the relationship between a mother’s milk were members of the core fecal genera in KE and ETR, but not microbiome and her infant’s fecal microbiome using both the ETU. Bacterial richness in feces from KE was greater than that Jaccard and Bray-Curtis distance metrics, the mother/infant from ETU, while diversity of milk from ETR and ETU were dyads’ samples were found to be more similar or tended to generally similar to that of KE. Additionally, milk collected from be more similar to each other than to all other combinations women in ETR contained relatively more Achromobacter and of mothers and infants (with Jaccard index: P = 0.0258; with Rhizobium than all other sites. While these differences are not Bray-Curtis dissimilarity: P = 0.0936). When evaluated within comprehensive among these cohorts, these examples illustrate each cohort, microbial communities of a mother’s milk and that even in close geographical proximity (KE to ETR and her infant’s feces were not more similar to each other than to ETU) and with genetic similarity (ETR and ETU, and GBR and all other combinations of mother/infant dyads (all P > 0.05 GBU), there are substantial differences between milk and infant for both matrices), except for in mother/infant samples from fecal microbial community structures that cannot be ignored, ETU in which the Bray-Curtis distance metric showed that particularly in the framework of recommendations for healthy bacterial communities between mothers and their infants were breastfeeding women/breastfed infants. more similar to each other (P = 0.0185) than to other random Our results are generally congruent with the limited number combinations within the cohort. of other studies of the infant fecal microbiome among similarly- aged infants. For example, Bifidobacterium, Streptococcus, and DISCUSSION the family Enterobacteriacae, which includes both Enterobacter and Escherichia/Shigella, were previously reported to be the most Data from this study support our hypotheses that: (1) the human abundant taxa in feces from Gambian infants (24). Our data show milk and infant fecal microbiomes vary among global cohorts; that infant feces from Gambian infants also contain substantial (2) there exists a small core group of bacteria common to amounts of Lactobacillus and Veillonella. Consistent with our milk across all cohorts (although some cohorts had additional results, Backhëd et al. (20) found Bacteroides, Bifidobacteria, taxa which composed their own unique core); (3) there exists Prevotella, Streptococcus, Veillonella, and Enterobacteriacae to be a small core group of bacteria common to infant feces across abundant in feces of Swedish infants. all cohorts (although some cohorts had their own cores more It is important to note that feces collected in ETR were different from the overall core); and (4) variation in the milk preserved with RNAlater, a method previously used for this microbiome is related to variation in the infant fecal microbiome purpose (48–50). However, RNAlater may introduce biases, such (although this was more apparent at the community level than as an increased observation of members of the Bacteroidetes the individual taxa level). Our hypothesis that fecal microbiomes phylum, and a decrease in the members of Actinobacteria of infants and the milk microbiomes of their mothers would relative to unpreserved control samples; this bias has been be more similar within a cohort than to other cohorts was also demonstrated in soil samples and fecal samples (51–55). Despite supported, though the variation among both women and infants this methodological difference, feces collected in ETR—at least within a cohort was only slightly less than the variation among with respect to phyla and the most abundant genera—were cohorts. This individual variation should be noted, because it similar to those collected at its closest neighboring site, ETU. is possible that milk and infant fecal microbiomes are tailored However, there were differences with respect to the core taxa, to a given environment. Additionally, the possibility exists that and samples from ETU formed a cluster with feces from KE and both the membership and structure of milk and fecal microbial PE, while ETR was an outgroup in this clade. Further work will communities may also be tailored to lifestyle and behavioral be needed to determine if the use of RNAlater does or does not factors associated with individual maternal/infant dyads (16, 46, influence bacterial communities identified in infant fecal samples 47). For this reason, these results highlight the need for additional collected across various cohorts. work comparing populations of women and infants globally if we Milk microbiomes characterized here were also relatively are ever to understand whether a “healthy” or “normal” human similar to those described in the literature previously. For milk or infant fecal microbiome exists. instance, Kumar et al. (17) reported that relative abundance

Frontiers in Nutrition | www.frontiersin.org 13 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

FIGURE 4 | Mean relative abundances of bacterial (A) phyla and (B) an aggregation of the 10 most-abundant genera in milk in each cohort. ETR, rural Ethiopia; ETU, urban Ethiopia; GBR, rural Gambia; GBU, urban Gambia; GN, Ghana; KE, Kenya; SP, Spain; SW, Sweden; PE, Peru; USC, California (United States); USW, Washington (United States). of Proteobacteria was highest in milk produced by women For this reason, results from ETR should be interpreted in South Africa; that produced by Finnish women was cautiously. For example, one notable difference we found was highest in Firmicutes; that of Chinese women had the highest the high richness and diversity in the ETR milk samples as Streptococcus; and that produced by Spanish women had the compared to the other sites. While this could be biologically highest Propionibacterium and Pseudomonas. In our study, milk relevant, this also could be an artifact of the method used for produced in ETR had the most Proteobacteria and Rhizobium, these samples. whereas that produced in PE had the highest level of Firmicutes, Important for further interpretation of the findings in this and that produced in SP the highest Propionibacterium. Despite study is the fact that, in addition to its microbial communities, these differences, most of the dominant taxa present in the study milk’s micronutrient and macronutrient compositions vary of Kumar and colleagues were also well represented in our study. globally. Whether milk’s microbiome and nutrient compositions Nonetheless, it remains unclear as to whether the differences in are related has not been studied, although data from our relative abundance between our present study and that of Kumar group clearly suggests that maternal nutrient intake (which can are genuine or due to methodologic differences. For example, be related to milk’s nutrient composition) is related to milk Kumar and colleagues used the same sequencing platform, but microbiome (56). One example particularly relevant to the data chose primers targeting the V4 hypervariable region of the 16S presented here is related to human milk oligosaccharides (HMO), rRNA gene, whereas our primers targeted the V1–V3 region. In which vary across populations—including those reported here this case, we also were only able to use the forward read, which (26, 57). Because HMO can act as prebiotics, it is possible that could have been a source of bias in elucidating the microbial variation in HMO profiles might drive variation in both milk genera present. and infant fecal microbiomes. This possibility will be explored Of note are the compositional differences in the ETR milk thoroughly in subsequent publications. We have also analyzed as compared to all other cohorts. Unlike all the others which the milk samples described in this report for their complex were frozen upon collection, these samples were chemically immune factor profiles, which also vary (25); as with HMO, preserved. The use of chemical preservatives at this site subsequent publications will evaluate potential relationships was necessary because freezing the samples was logistically among milk’s immune factors, the milk microbiome, and difficult. While Milk Preservation Solution performed well the infant fecal microbiome. While the etiology of these (as compared to other preservatives) in a test of utility in population-level differences has not been completely elucidated, preserving bacterial DNA for microbial analysis (28), the the potential importance of these multi-faceted differences methods used for extracting the ETR samples are different in shaping milk’s microbial communities should continue to than those employed on all other milk samples in this study. be evaluated.

Frontiers in Nutrition | www.frontiersin.org 14 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study ton bc abc cd a b a a c b b b b ab b b ad b a b 39 3.3 3.8 0.7 0.9 0.7 0.3 0.9 0.5 0.2 0.2 0.2 0.2 0.5 0.3 0.2 0.1 0.2 0.2 0.2 0.2 0.1 0.2 0.1 0.1 0.1 0.1 0.2 0.0 0.0 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 4.1 5.6 1.4 6.4 2.9 1.1 1.3 0.9 1.2 2.2 1.6 0.3 0.8 1.1 0.4 1.0 0.2 26.5 30.4 bc abc cde ab b cd ab c ab b bc b ab b b abcd b cd b 19 4.6 4.9 0.8 2.3 0.4 0.9 3.8 0.3 0.5 0.3 0.4 0.2 0.3 0.6 0.4 0.1 0.3 1.0 0.2 0.2 1.1 0.2 0.9 0.1 0.1 0.1 0.3 0.2 0.7 0.1 0.1 0.6 0.1 0.3 0.1 0.4 0.1 0.0 0.2 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 3.2 3.7 1.4 1.3 2.1 1.0 1.8 0.8 1.3 2.3 1.7 0.3 0.8 0.5 0.4 0.3 0.2 25.7 26.0 bce a cdef a b bcd ab c ab bc b b a b b abc b ab b 23 4.5 5.6 0.8 1.3 0.3 1.0 3.7 0.4 0.5 0.3 0.4 0.2 0.3 0.5 0.7 0.1 0.2 1.3 0.2 0.2 0.8 0.2 0.8 0.1 0.1 0.1 0.3 0.2 0.7 0.2 0.1 0.5 0.1 0.3 0.1 0.4 0.0 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 3.9 6.6 1.2 2.0 2.0 1.1 1.7 1.0 1.2 2.2 2.9 0.3 0.9 0.7 0.5 0.7 0.2 29.5 40.7 ab abc de a b bc a c ab b ab b ab b b abc b abcd b 40 3.8 3.5 0.4 0.7 0.8 4.3 0.3 0.4 0.5 0.2 0.3 0.2 0.3 0.4 0.4 0.2 1.0 0.1 0.2 0.7 0.2 0.9 0.2 0.1 0.1 0.3 0.2 0.7 0.1 0.1 0.4 0.1 0.4 0.1 0.3 0.1 0.0 0.1 0.0 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 2.5 4.3 1.4 2.6 3.1 1.2 1.4 1.0 1.6 2.1 2.0 0.3 0.8 0.6 0.4 0.5 0.2 36.5 27.3 ae abc cd ab b bc a c a b b b a b b bc b bcd b 43 3.9 3.5 0.7 0.7 0.2 0.8 4.1 0.4 0.6 0.2 0.5 0.2 0.3 0.4 0.5 0.1 0.2 1.0 0.2 0.1 0.9 0.2 0.9 0.1 0.1 0.1 0.3 0.1 0.2 0.7 0.1 0.5 0.1 0.3 0.1 0.3 0.0 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 4.2 4.1 1.3 2.7 3.3 1.2 2.8 1.0 1.4 2.1 2.8 0.3 0.8 0.5 0.6 0.4 0.2 49.6 29.6 b abc bcd a b bc ab abc ab ab ab b ab b b abc ab abcd b 42 3.5 3.4 0.8 0.9 0.2 1.2 4.1 0.5 0.4 0.4 0.3 0.2 0.4 0.5 0.3 0.1 0.2 0.9 0.2 0.1 0.7 0.2 0.8 0.1 0.1 0.1 0.3 0.1 0.2 0.8 0.1 0.4 0.1 0.3 0.1 0.4 0.0 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 4.8 5.8 1.2 2.9 2.4 2.2 1.6 1.3 2.0 2.3 1.6 0.3 1.0 0.6 0.6 0.5 0.2 31.5 27.7 0.05). d abc e b b d b c b ab b a b b b bc b cd b > 37 3.0 1.4 0.3 0.4 0.2 0.2 1.5 6.7 0.3 0.2 0.2 0.3 0.2 1.3 0.2 0.1 0.2 0.4 1.0 0.1 0.1 0.6 0.3 0.8 0.1 0.1 0.0 0.1 0.4 0.2 0.7 0.1 0.5 0.1 0.4 0 0.3 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 8.7 1.8 2.0 1.2 1.2 1.3 1.1 1.3 1.4 1.0 7.0 1.1 0.3 0.8 0.4 0.4 0.3 0.2 20.6 bc bc bc ab b ab a abc b ab b b ab b ab ab b bcd b a urban Ghana Kenya Peru Spain Sweden US California US Washing 38 3.3 3.0 0.9 0.7 0.3 0.6 0.8 8.3 0.6 0.4 0.2 0.2 0.3 0.4 0.3 0.1 0.3 0.2 1.8 0.1 0.2 0.6 0.2 1.7 0.1 0.1 0.0 0.1 0.3 0.2 0.8 0.1 0.4 0.1 0.6 0.1 0.2 0.0 0.2 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n different from each other (P 5.5 3.6 1.5 3.6 3.1 2.3 1.1 1.3 1.5 2.2 1.4 0.3 1.5 0.8 0.5 0.4 0.2 26.0 20.5 SEM. b c bcd ab b ab a bc b ab ab b ab b a abc b abc b ± 39 29 bacterial genera in milk. 3.5 2.6 0.8 0.7 0.3 0.6 0.8 4.3 0.6 0.3 0.2 0.3 0.3 0.4 0.3 0.1 0.4 0.2 0.9 0.1 0.1 0.8 0.2 0.9 0.1 0.1 0.0 0.1 0.3 0.2 0.7 0.1 0.4 0.1 0.3 0.1 0.3 0.0 0.2 s = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 4.8 3.8 1.5 3.6 3.5 1.6 1.1 1.6 1.6 2.1 1.4 0.3 2.4 0.6 0.5 0.6 0.2 29.5 17.5 bc ab abf a b bcd a ab b ac ab b ab b ab bc ab bcd b 34 3.5 4.3 1.4 0.8 0.3 0.9 4.3 0.3 0.6 0.6 0.2 0.4 0.3 0.5 0.3 0.1 0.2 0.9 0.2 0.1 0.7 0.2 0.8 0.1 0.1 0.1 0.1 0.3 0.2 0.7 0.0 0.1 0.4 0.1 0.3 0.1 0.3 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 8.8 4.4 1.3 1.8 3.0 3.3 1.2 2.4 1.7 2.1 1.4 0.3 1.2 0.5 0.7 0.3 0.2 25.9 34.2 cd abc a ab a d ab a ab a ac b ab a b c a a d 40 2.2 3.0 1.5 0.5 2.2 1.3 4.7 0.2 0.3 0.7 0.4 0.4 0.6 0.4 0.3 0.7 0.2 0.9 0.2 0.2 0.6 0.2 0.8 0.1 0.2 0 0.1 0.3 0.2 0.7 0.1 0.1 0.4 0.1 0.3 0.1 0.3 0.0 0.1 = ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n 3.0 7.0 0.9 1.7 4.4 2.0 2.5 3.3 2.2 1.5 8.1 1.2 0.9 0.9 1.1 0.3 1.1 0.2 0.5 0.7 0.8 0.4 0.3 0.3 0.1 15.3 21.5 11.0 22.5 0.05) as determined by one-way ANOVA; values sharing a letter are not 1 1 ≤ 1 1 1 1 1 1 1 1 Effect of cohort on relative abundances (%) of the aggregate 1 1 1 1 1 1 1 1 1 Effect of cohort (P TABLE 5 | GenusStreptococcus Ethiopia rural Ethiopia urban The Gambia rural The Gambi Values are model estimates based on a beta distribution and represent mean 1 Staphylococcus Corynebacterium 1 Propionibacterium Rhizobium Lactobacillus Dyella Rothia Kocuria Veillonella Bifidobacterium Acinetobacter Klebsiella Gemella Achromobacter Escherichia/Shigella Bacillus Stenotrophomonas Enterococcus Janthinobacterium Anaerococcus Acidocella Tatumella Psychrobacter Enterobacter Bacteroides Pseudomonas Chryseobacterium Clostridium sensu stricto 18

Frontiers in Nutrition | www.frontiersin.org 15 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

FIGURE 5 | Spearman rank correlations between an aggregation of the 10 most-abundant bacterial genera in infant feces and an aggregation of the 10

most-abundant bacterial genera in milk. Stars indicate P < 0.01 and rs < −0.3 or rs > 0.3.

Importantly, our data provide evidence of relationships fecal bacterial communities collected at the same time. With between the milk microbiome and the fecal microbiome of the exception of GN, more taxa were identified in milk than breastfed infants. For example, the relative abundance of infant feces here as well. However, except for KE where bacterial Lactobacillus in milk was positively correlated with the relative richness of infant feces was positively correlated with that of abundance of Lactobacillus in infant feces. Given the fact milk, we found no correlations between the richness of milk and that this genus was also determined to be the primary factor richness of infant feces in the populations we studied. that distinguished the first and second canonical axes in our There are several important limitations that should be multivariate correlation analyses, it is plausible that human considered when interpreting the findings reported here. As milk may be an important source of Lactobacillus for the previously discussed, we chemically preserved samples collected developing infant’s GI tract, as has been demonstrated through only in ETR. In addition, for practical and logistic reasons, culture-dependent analyses (3–5). Future work focusing on we also used a combination of methods for collecting the the species and strains of Lactobacillus at both body sites is milk (electric pump vs. hand expression). The impact of these needed to understand these complex communities. Our data methodological differences on milk’s microbial composition is support a handful of studies published previously on this unknown. Additionally, limited cohort sample sizes (particularly topic. For example, Murphy et al. (12) collected milk and with respect to USC), restricts the ability to conduct some infant fecal samples from 10 mother-infant pairs at 1, 3, 6, statistical analyses. Furthermore, because this study is the first of and 12 wk postpartum. Approximately 70–88% of the genera its kind to this scale, we do not know if our results are comparable identified in infant feces were also identified in human milk. to other studies on milk and fecal microbial compositions, The most abundant genera in infant feces were Streptococcus, particularly as it relates to methodological differences. For Escherichia/Shigella, Bifidobacterium, and Veillonella, which were example, some genera, such as Bifidobacterium, Ruminococcus, the 4 most abundant genera in infant fecal samples in our and Coprococcus, are present in higher relative abundances study. Murphy and colleagues also observed that human milk when mechanical lysis (via bead-beating) is used (58) due bacterial communities exhibited greater diversity than infant to the difficulty in cell membrane disruption for these taxa.

Frontiers in Nutrition | www.frontiersin.org 16 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

FIGURE 6 | Canonical correlations between aggregations of the 10 most-abundant bacterial genera in milk (x-axes) and aggregations of the 10 most-abundant bacterial genera in infant feces (y-axes) in each cohort. The overall correlation is plotted in the upper-left panel; individual cohorts are illustrated in subsequent plots.

It is important to note that characterization and analyses of into these findings, such as antibiotic usage, infant age, parity, the bacteria at a level lower than their genera here were not infant sex, and exclusive breastfeeding status. These factors are performed. Future studies should evaluate these communities likely important to understanding what are normal for milk at the species (and perhaps sub-species) level. The selection of and infant fecal microbial community structures. The goal of a hypervariable region to target for 16S rRNA analysis is also the current study was to present the microbial communities a source of potential bias in our study (59). We chose V1– across a broad range of populations, and subsequent studies V3 because it has been previously used in our laboratory to will focus on parsing the relationships among these factors categorize and classify the bacterial community structure of both and the milk and infant fecal microbiomes. For these findings milk and infant feces (16, 30, 56). However, future studies should to have impact, however, researchers must strive to better be designed to determine if this hypervariable region is optimal understand the genesis of microbial differences within and for this application. Additionally, one important limitation to among cohorts, and if this variation is related to maternal and 16S-based sequencing techniques is the inability to distinguish infant health. live bacteria from dead bacteria or residual bacterial DNA that In conclusion, this study is the largest of its kind to use may be present in a sample. As more is understood about standardized methodologies to characterize and compare the the microbial communities in various populations globally and milk and infant fecal microbiomes of maternal/infant dyads at various body sites, more targeted, culture- or RNA-based worldwide. Substantial differences in both the composition and approaches can be used in conjunction with 16S sequencing to relative abundance of specific taxa were present among cohorts. more thoroughly address these questions of live contributors to We also found substantial variation among women/infants the milk and infant fecal microbiomes. within a cohort, suggesting that environment alone does not Nonetheless, this research represents the largest, cross- drive variation in milk and fecal microbial community structure. cultural, international study to date using standardized methods Additionally, relationships present among the genera in milk and of collection and analysis for characterizing the microbial the genera in infant feces which vary among cohorts suggest that compositions of human milk and infant feces. Data from this milk may be tailored not only to infants in a given environment, study clearly indicate that what is “normal” in terms of milk and but also specifically to the needs of individuals. As such, we fecal microbiomes of healthy breastfeeding mothers and their conclude that what is “normal” in terms of the fecal microbiome infants, respectively, varies around the world. In addition, we of healthy infants and milk microbiome of healthy lactating provide compelling evidence that the milk microbiome might, at women varies by culture and/or location. Further, we posit that least in part, play a role in shaping the infant’s GI microbiome. bacterial compositions of human milk and feces of breastfed It is likely that there are many additional factors that play infants are likely specific to a dyad within a culture and location.

Frontiers in Nutrition | www.frontiersin.org 17 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

Future studies should evaluate the bacterial species and sub- in Ghana; Katherine Flores (Washington State University), species encompassed in these genera, their functionality, and Dubale Gebeyehu (Hawassa University), Haile Belachew whether or not variation is related to maternal and infant health. (Hawassa University), and Birhanu Sintayehu for planning, logistics, recruiting, and data collection and the administration ETHICS STATEMENT and staff at Adare Hospital in Hawassa for assistance with logistics in Ethiopia; Catherine O Sarange (Egerton University) This study was carried out in accordance with the for field supervision and logistics planning and Milka W. recommendations of the Washington State University Churuge and Minne M Gachau for recruiting, questionnaire Institutional Review Board and the ethics approval committees administration, and taking anthropometric measurements at each participating location with informed consent from all in Kenya; Gisella Barbagelatta (Instituto de Investigación subjects. All subjects gave informed consent in accordance with Nutricional) for field supervision and logistics planning, the Declaration of Helsinki. The protocol was approved by Patricia Calderon (Instituto de Investigación Nutricional) the Washington State University Institutional Review Board for recruiting, questionnaire administration, and taking (approval #13264). anthropometric measurements, and Roxana Barrutia (Instituto de Investigación Nutricional) for the management and shipping AUTHOR CONTRIBUTIONS of samples in Peru; Leonides Fernandez, and Irene Espinosa (Complutense University of Madrid) for technical assistance, MKM, CM, MAM, JW, JF, DS, LK, JR, LB, RP, and AP designed expertise, and review of the manuscript and M Ángeles Checa the study. KL, JZ, EB, CM, EK-M, EK, DK, SM, SEM, GO, (Zaragoza, Spain), Katalina Legarra (Guernica, Spain), and Julia CG-C, EJ, and LR collected the samples, KL, EB, and JZ Mínguez (Huesca, Spain) for participation in the collection performed the laboratory analyses, KL, WP, and JW performed of samples in Spain; Kirsti Kaski and Maije Sjöstrand (both the data analyses, KL and MKM wrote the manuscript. All Helsingborg Hospital) for participation in the collection of authors read, contributed to, and approved the final manuscript. samples, questionnaire administration, and anthropometric measurements in Sweden; Renee Bridge and Kara Sunderland FUNDING (both University of California, San Diego); Janae Carrothers and Shelby Hix (Washington State University) for logistics This study was supported by the National Science Foundation planning, recruiting, questionnaire administration, sample (award 1344288), the Ministry of Economy and Competitiveness, collection, and taking anthropometric measurements in Spain (project AGL2013-4190-P), and the European California and Washington; Romana J. Hyde and Morgan Commission [grant 624773 (FP-7-PEOPLE-2013-IEF)]. Sterile, M. Potton for their help in sample analysis; Glenn Miller single-use milk-collection kits were generously provided by (Washington State University) for his expertise and critical Medela Inc. This work was supported in part by NIH COBRE logistic help that were needed for shipping samples and Phase III grant P30GM103324. supplies worldwide; and Ryan Pace (University of Idaho) for his assistance in bioinformatics and analyses and review of ACKNOWLEDGMENTS the manuscript. We sincerely thank Andrew Doel (Medical Research Council SUPPLEMENTARY MATERIAL Unit, The Gambia) for field supervision and logistics planning and Alansan Sey for questionnaire administration and taking The Supplementary Material for this article can be found anthropometric measurements in The Gambia; Jane Odei online at: https://www.frontiersin.org/articles/10.3389/fnut.2019. (University of Ghana) for supervising field data collection 00045/full#supplementary-material

REFERENCES role in the colonization of the infant gut. J Appl Microbiol. (2007) 103:2638–44. doi: 10.1111/j.1365-2672.2007.03497.x 1. Fernández L, Langa S, Martín V, Maldonado A, Jiménez E, Martín 5. Martín V, Maldonado-Barragán A, Moles L, Rodriguez-Baños M, Campo R, et al. The human milk microbiota: origin and potential roles in RD, Fernández L, et al. Sharing of bacterial strains between breast milk health and disease. Pharmacol Res. (2013) 69:1–10. doi: 10.1016/j.phrs.201 and infant feces. J Hum Lact. (2012) 28:36–44. doi: 10.1177/08903344114 2.09.001 24729 2. Milani C, Duranti S, Bottacini F, Casey E, Turroni F, Mahony J, et al. The first 6. Stiemsma LT, Michels KB. The role of the microbiome in the developmental microbial colonizers of the human gut: composition, activities, and health origins of health and disease. Pediatrics. (2018) 141:e20172437. implications of the infant gut microbiota. Microbiol Mol Biol Rev. (2017) doi: 10.1542/peds.2017-2437 81:e00036–17. doi: 10.1128/MMBR.00036-17 7. Hunt KM, Foster JA, Forney LJ, Schütte UM, Beck DL, Abdo Z, 3. Martín R, Langa S, Reviriego C, Jimínez E, Marín ML, Xaus J, et al. Human et al. Characterization of the diversity and temporal stability of milk is a source of lactic acid bacteria for the infant gut. J Pediatr. (2003) bacterial communities in human milk. PLoS ONE. (2011) 6:e21313. 143:754–8. doi: 10.1016/j.jpeds.2003.09.028 doi: 10.1371/journal.pone.0021313 4. Martín R, Heilig GH, Zoetendal EG, Smidt H, Rodríguez JM. Diversity of the 8. Cabrera-Rubio R, Collado MC, Laitinen K, Salminen S, Isolauri E, Mira Lactobacillus group in breast milk and vagina of healthy women and potential A. The human milk microbiome changes over lactation and is shaped by

Frontiers in Nutrition | www.frontiersin.org 18 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

maternal weight and mode of delivery. Am J Clin Nutr. (2012) 96:544–51. 27. World Health Organization. Child Growth Standards: Anthro and Macros. R doi: 10.3945/ajcn.112.037382 (version 3.2.2). (2011). Available online at: http://www.who.int/childgrowth/ 9. Davé V, Street K, Francis S, Bradman A, Riley L, Eskenazi B, et al. software/en/ Bacterial microbiome of breast milk and child saliva from low-income 28. Lackey KA, Williams JE, Price WJ, Carrothers JM, Brooker SL, Shafii B, Mexican-American women and children. Pediatr Res. (2016) 79:846–54. et al. Comparison of commercially-available preservatives for maintaining doi: 10.1038/pr.2016.9 the integrity of bacterial DNA in human milk. J Microbiol Methods. (2017) 10. Li SW, Watanabe K, Hsu CC, Chao SH, Yang ZH, Lin YJ, et al. 141:73–81. doi: 10.1016/j.mimet.2017.08.002 Bacterial composition and diversity in breast milk samples from mothers 29. Frank JA, Reich CI, Sharma S, Weisbaum JS, Wilson BA, Olsen GJ. living in Taiwan and mainland China. Front Microbiol. (2017) 8:965. Critical evaluation of two primers commonly used for amplification of doi: 10.3389/fmicb.2017.00965 bacterial 16S rRNA genes. Appl Environ Microbiol. (2008) 74:2461–70. 11. Urbaniak C, Angelini M, Gloor GB, Reid G. Human milk microbiota profiles doi: 10.1128/AEM.02272-07 in relation to birthing method, gestation and infant gender. Microbiome. 30. Carrothers JM, York MA, Brooker SL, Lackey KA, Williams JE, Shafii B, (2016) 4:1. doi: 10.1186/s40168-015-0145-y et al. Fecal microbial community structure is stable over time and related to 12. Murphy K, Curley D, O’Callaghan TF, O’Shea CA, Dempsey EM, O’Toole variation in macronutrient and micronutrient intakes in lactating women. J PW, et al. The composition of human milk and infant faecal microbiota Nutr. (2015) 145:2379–88. doi: 10.3945/jn.115.211110 over the first three months of life: a pilot study. Sci Rep. (2017) 7:40597. 31. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. doi: 10.1038/srep40597 DADA2: High-resolution sample inference from Illumina amplicon data. Nat 13. Vaidya YH, Patel SH, Patel RJ, Pandit RJ, Joshi CG, Kunjadiac AP. Human Methods. (2016) 13:581–3. doi: 10.1038/nmeth.3869 milk microbiome in urban and rural populations of India. Meta Gene. (2017) 32. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid 13:13–22. doi: 10.1016/j.mgene.2017.04.001 assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ 14. Pannaraj PS, Li F, Cerini C, Bender JM, Yang S, Rollie A, et al. Microbiol. (2007) 73:5261–7. doi: 10.1128/AEM.00062-07 Association between breast milk bacterial communities and establishment and 33. Callahan BJ. The Silva taxonomic training set formatted for DADA2 [Data development of the infant gut microbiome. JAMA Pediatr. (2017) 171:647–54. set]. Zenodo. (2016) doi: 10.5281/zenodo.158958 doi: 10.1001/jamapediatrics.2017.0378 34. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The 15. Patel SH, Vaidya YH, Patel RJ, Pandit RJ, Joshi CG, Kunjadiya AP. Culture SILVA ribosomal RNA gene database project: improved data processing independent assessment of human milk microbial community in lactational and web-based tools. Nucleic Acids Res. (2013) 41(Database issue):D590-6. mastitis. Sci Rep. (2017) 7:7804. doi: 10.1038/s41598-017-08451-7 doi: 10.1093/nar/gks.1219 16. Meehan CL, Lackey KA, Hagen EH, Williams JE, Roulette J, Helfrecht C, 35. Yilmaz P, Parfrey LW, Yarza P, Gerken J, Pruesse E, Quast C, et al. The SILVA et al. Social networks, cooperative breeding, and the human milk microbiome and “All-species Living Tree Project (LTP)” taxonomic frameworks. Nucleic (HMM). Am J Hum Biol. (2018) 30:e23131. doi: 10.1002/ajhb.23131 Acids Res. (2014) 42(Database issue):D643–8. doi: 10.1093/nar/gkt1209 17. Kumar H, du Toit E, Kulkarni A, Aakko J, Linderborg KM, Zhang 36. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive Y, et al. Distinct patterns in human milk microbiota and fatty acid analysis and graphics of microbiome census data. PLoS ONE. (2013) 8:e61217. profiles across specific geographic locations. Front Microbiol. (2016) 7:1619. doi: 10.1371/journal.pone.0061217 doi: 10.3389/fmicb.2016.01619 37. Whittaker RH. Evolution and measurement of species diversity. Taxon. (1972) 18. Yatsunenko T, Rey FE, Manary MJ, Trehan I, Dominguez-Bello MG, Contreras 21:213–51. doi: 10.2307/1218190 M, et al. Human gut microbiome viewed across age and geography. Nature. 38. Morris EK, Caruso T, Buscot F, Fischer M, Hancock C, Maier TS, et al. (2012) 486:222–7. doi: 10.1038/nature11053 Choosing and using diversity indices: insights for ecological applications 19. Azad MB, Konya T, Maughan H, Guttman DS, Field CJ, Chari RS, et al. Gut from the German Biodiversity Exploratories. Ecol Evol. (2014) 4:3514–24. microbiota of healthy Canadian infants: profiles by mode of delivery and doi: 10.1002/ece3.1155 infant diet at 4 months. CMAJ. (2013) 85:385–94. doi: 10.1503/cmaj.121189 39. Shannon CE. A mathematical theory of communication. Bell Syst Tech J. 20. Bäckhed F, Roswall J, Peng Y, Feng Q, Jia H, Kovatcheva-Datchary P, et al. (1948) 27:379–423. doi: 10.1002/j.1538-7305.1948.tb01338.x Dynamics and stabilization of the human gut microbiome during the first 40. McCune B, Grace JB. Analysis of Ecological Communities. Gleneden Beach year of life. Cell Host Microbe. (2015) 17:852. doi: 10.1016/j.chom.2015. Oregon: MjM and Software Design, (2002). 05.012 41. Fisher RA, Corbet AS, Williams CB. The relation between the number of 21. Rodríguez JM, Murphy K, Stanton C, Ross RP, Kober OI, Juge N, et al. The species and the number of individuals in a random sample of an animal composition of the gut microbiota throughout life, with an emphasis on population. J Anim Ecol. (1943) 12:42–58. doi: 10.2307/1411 early life. Microb Ecol Health Dis. (2015) 26:26050. doi: 10.3402/mehd.v26. 42. R Core Team. R: A Language and Environment for Statistical Computing. 26050 Vienna: R Foundation for Statistical Computing, (2016). Available online 22. Martin R, Makino H, Cetinyurek Yavuz A, Ben-Amor K, Roelofs M, Ishikawa at: https://www.R-project.org/ E, et al. Early-life events, including mode of delivery and type of feeding, 43. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, et al. siblings and gender, shape the developing gut microbiota. PLoS ONE. (2016) vegan: Community Ecology Package. R Package Version 2.5-2. (2018). Available 11:e0158498. doi: 10.1371/journal.pone.0158498 online at: https://CRAN.R-project.org/package=vegan 23. Grze´skowiak Ł, Collado MC, Mangani C, Maleta K, Laitinen K, Ashorn 44. Wickham H. ggplot2: Elegant Graphics for Data Analysis. New York, NY: P, et al. Distinct gut microbiota in southeastern African and northern Springer-Verlag, (2009). European infants. J Pediatr Gastroenterol Nutr. (2012) 54:812–6. 45. Warnes GR, Bolker B, Bonebakker L, Gentleman R, Huber W, Liaw A, et al. doi: 10.1097/MPG.0b013e318249039c gplots: Various R Programming Tools for Plotting Data. R Package Version 24. Davis JC, Lewis ZT, Krishnan S, Bernstein RM, Moore SE, Prentice AM, 3.0.1. (2016). Available online at: https://CRAN.R-project.org/package=gplots et al. Growth and morbidity of Gambian infants are influenced by maternal 46. McGuire MK, Meehan CL, Brooker SL, Williams JE, Foster JA, and McGuire milk oligosaccharides and infant gut microbiota. Sci Rep. (2017) 7:40466. MA. An evolutionary, biosocial perspective on variation in human milk doi: 10.1038/srep40466 microbes and oligosaccharides: An example of eco-homeorhesis? In: McGuire 25. Ruiz L, Espinosa-Martos I, García-Carral C, Manzano S, McGuire MK, MK, McGuire MA, Bode L, editors. Prebiotics and Probiotics in Human Milk. Meehan CL, et al. What’s Normal? Immune profiling of human milk from Amsterdam: Elselvier, Inc. (2017). healthy women living in different geographical and socioeconomic settings. 47. Ojo-Okunola A, Nicol M, du Toit E. Human breast milk bacteriome in health Front Immunol. (2017) 8:696. doi: 10.3389/fimmu.2017.00696 and disease. Nutrients. (2018) 10:E1643. doi: 10.3390/nu10111643 26. McGuire MK, Meehan CL, McGuire MA, Williams JE, Foster J, Sellen DW, 48. De Filippo C, Cavalieri D, Di Paola M, Ramazzotti M, Poullet JB, Massart S, et al. What’s normal? Oligosaccharide concentrations and profiles in milk et al. Impact of diet in shaping gut microbiota revealed by a comparative study produced by healthy women vary geographically. Am J Clin Nutr. (2017) in children from Europe and rural Africa. Proc Natl Acad Sci USA. (2010) 105:1086–100. doi: 10.3945/ajcn.116.139980 107:14691–6. doi: 10.1073/pnas.1005963107

Frontiers in Nutrition | www.frontiersin.org 19 April 2019 | Volume 6 | Article 45 Lackey et al. The INSPIRE Study

49. VlckovᡠK, Mrázek J, Kopecnýˇ J, and Petrželková KJ. Evaluation of different and gut in moderately preterm infants. Front Microbiol. (2018) 9:2512. storage methods to characterize the fecal bacterial communities of captive doi: 10.3389/fmicb.2018.02512 western lowland gorillas (Gorilla gorilla). J Microbiol Methods. (2012) 91:45– 63. Chen PW, Lin YL, Huang MS. Profiles of commensal and 51. doi: 10.1016/j.mimet.2012.07.015 opportunistic bacteria in human milk from healthy donors in Taiwan. 50. Flores R, Shi J, Yu G, Ma B, Ravel J, Goedert JJ, et al. Collection media J Food Drug Anal. (2018) 26:1235–44. doi: 10.1016/j.jfda.2018. and delayed freezing effects on microbial composition of human stool. 03.004 Microbiome. (2015) 3:33. doi: 10.1186/s40168-015-0092-7 64. Ward TL, Hosid S, Ioshikhes I, Altosaar I. Human milk metagenome: 51. Dominianni C, Wu J, Hayes RB, Ahn J. Comparison of methods for a functional capacity analysis. BMC Microbiol. (2013) 13:116. fecal microbiome biospecimen collection. BMC Microbiol. (2014) 14:103. doi: 10.1186/1471-2180-13-116 doi: 10.1186/1471-2180-14-103 65. Boix-Amorós A, Collado MC, Mira A. Relationship between milk microbiota, 52. McCarthy A, Chiang E, Schmidt ML, Denef VJ. RNA preservation bacterial load, macronutrients, and human cells during lactation. Front agents and nucleic acid extraction method bias perceived Microbiol. (2016) 7:492. doi: 10.3389/fmicb.2016.00492 bacterial community composition. PLoS ONE. (2015) 10:e0121659. 66. Cabrera-Rubio R, Mira-Pascual L, Mira A, Collado MC. Impact of mode doi: 10.1371/journal.pone.0121659 of delivery on the milk microbiota composition of healthy women. 53. Gorzelak MA, Gill SK, Tasnim N, Ahmadi-Vand Z, Jay M, Gibson J Dev Orig Health Dis. (2016) 7:54–60. doi: 10.1017/S20401744150 DL. Methods for improving human gut microbiome data by reducing 01397 variability through sample processing and storage of stool. PLoS ONE. (2015) 67. Tuominen H, Rautava S, Collado MC, Syrjänen S, Rautava J. HPV infection 10:e0134802. doi: 10.1371/journal.pone.0134802 and bacterial microbiota in breast milk and infant oral mucosa. PLoS ONE. 54. Song SJ, Amir A, Metcalf JL, Amato KR, Xu ZZ, Humphrey G, (2018) 13:e0207016. doi: 10.1371/journal.pone.0207016 et al. Preservation methods differ in fecal microbiome stability, 68. Urbaniak C, McMillan A, Angelini M, Gloor GB, Sumarah M, Burton JP, et al. affecting suitability for field studies. mSystems. (2016) 1:e00021-16. Effect of chemotherapy on the microbiota and metabolome of human milk, a doi: 10.1128/mSystems.00021-16 case report. Microbiome. (2014) 2:24. doi: 10.1186/2049-2618-2-24 55. Choo JM, Leong LE, Rogers GB. Sample storage conditions significantly 69. Cabrera-Rubio R, Kunz C, Rudloff S, Garcia-Mantrana I, Crehuá- influence faecal microbiome profiles. Sci Rep. (2015) 5:16350. Gaudiza E, Martinez-Costa C, et al. Association of maternal secretor doi: 10.1038/srep16350 status and human milk oligosaccharides with milk microbiota: an 56. Williams JE, Carrothers JM, Lackey KA, Beatty NF, York MA, Brooker observational pilot study. J Pediatr Gastroenterol Nutr. (2018) 68:256–63. SL, et al. Human milk microbial community structure is relatively stable doi: 10.1097/MPG.0000000000002216 and related to variations in macronutrient and micronutrient intakes in 70. Biagi E, Quercia S, Aceti A, Beghetti I, Rampelli S, Turroni S, healthy lactating women. J Nutr. (2017) 147:1739–48. doi: 10.3945/jn.117. et al. The bacterial ecosystem of mother’s milk and infant’s mouth 248864 and gut. Front Microbiol. (2017) 8:1214. doi: 10.3389/fmicb.201 57. Erney RM, Malone WT, Skelding MB, Marcon AA, Kleman-Leyer KM, 7.01214 O’Ryan ML, et al. Variability of human milk neutral oligosaccharides in a diverse population. J Pediatr Gastroenterol Nutr. (2000) 30:181–92. Conflict of Interest Statement: CG-C and EJ are employed by company doi: 10.1097/00005176-200002000-00016 Probisearch. 58. Lim MY, Song EJ, Kim SH, Lee J, Nam YD. Comparison of DNA extraction methods for human gut microbial community profiling. Syst Appl Microbiol. The remaining authors declare that the research was conducted in the absence of (2018) 41:151–7. doi: 10.1016/j.syapm.2017.11.008 any commercial or financial relationships that could be construed as a potential 59. Tremblay J, Singh K, Fern A, Kirton ES, He S, Woyke T, et al. Primer and conflict of interest. platform effects on 16S rRNA tag sequencing. Front Microbiol. (2015) 6:771. doi: 10.3389/fmicb.2015.00771 Copyright © 2019 Lackey, Williams, Meehan, Zachek, Benda, Price, Foster, Sellen, 60. Jiménez E, de Andrés J, Manrique M, Pareja-Tobes P, Tobes R, Martínez- Kamau-Mbuthia, Kamundia, Mbugua, Moore, Prentice, K., Kvist, Otoo, García- Blanch JF, et al. Metagenomic analysis of milk of healthy and mastitis-suffering Carral, Jiménez, Ruiz, Rodríguez, Pareja, Bode, McGuire and McGuire. This is an women. J Hum Lact. (2015) 31:406–15. doi: 10.1177/0890334415585078 open-access article distributed under the terms of the Creative Commons Attribution 61. Jost T, Lacroix C, Braegger C, Chassard C. Assessment of bacterial diversity in License (CC BY). The use, distribution or reproduction in other forums is permitted, breast milk using culture-dependent and culture-independent approaches. Br provided the original author(s) and the copyright owner(s) are credited and that the J Nutr. (2013) 110:1253–62. doi: 10.1017/S0007114513000597 original publication in this journal is cited, in accordance with accepted academic 62. Biagi E, Aceti A, Quercia S, Beghetti I, Rampelli S, Turroni S, et al. practice. No use, distribution or reproduction is permitted which does not comply Microbial community dynamics in mother’s milk and infant’s mouth with these terms.

Frontiers in Nutrition | www.frontiersin.org 20 April 2019 | Volume 6 | Article 45